Table of Contents


This manual is for GNU Autoconf (version 2.59, 11 July 2004), a package for creating scripts to configure source code packages using templates and an M4 macro package.

Copyright © 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001, 2002, 2003 Free Software Foundation, Inc.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.1 or any later version published by the Free Software Foundation; with no Invariant Sections, with the Front-Cover texts being “A GNU Manual,” and with the Back-Cover Texts as in (a) below. A copy of the license is included in the section entitled “GNU Free Documentation License.”

(a) The FSF's Back-Cover Text is: “You have freedom to copy and modify this GNU Manual, like GNU software. Copies published by the Free Software Foundation raise funds for GNU development.”

1 Introduction

A physicist, an engineer, and a computer scientist were discussing the

nature of God. “Surely a Physicist,” said the physicist, “because

early in the Creation, God made Light; and you know, Maxwell's

equations, the dual nature of electromagnetic waves, the relativistic

consequences...” “An Engineer!,” said the engineer, “because

before making Light, God split the Chaos into Land and Water; it takes a

hell of an engineer to handle that big amount of mud, and orderly

separation of solids from liquids...” The computer scientist

shouted: “And the Chaos, where do you think it was coming from, hmm?”


Autoconf is a tool for producing shell scripts that automatically configure software source code packages to adapt to many kinds of unix-like systems. The configuration scripts produced by Autoconf are independent of Autoconf when they are run, so their users do not need to have Autoconf.

The configuration scripts produced by Autoconf require no manual user intervention when run; they do not normally even need an argument specifying the system type. Instead, they individually test for the presence of each feature that the software package they are for might need. (Before each check, they print a one-line message stating what they are checking for, so the user doesn't get too bored while waiting for the script to finish.) As a result, they deal well with systems that are hybrids or customized from the more common unix variants. There is no need to maintain files that list the features supported by each release of each variant of unix.

For each software package that Autoconf is used with, it creates a configuration script from a template file that lists the system features that the package needs or can use. After the shell code to recognize and respond to a system feature has been written, Autoconf allows it to be shared by many software packages that can use (or need) that feature. If it later turns out that the shell code needs adjustment for some reason, it needs to be changed in only one place; all of the configuration scripts can be regenerated automatically to take advantage of the updated code.

The Metaconfig package is similar in purpose to Autoconf, but the scripts it produces require manual user intervention, which is quite inconvenient when configuring large source trees. Unlike Metaconfig scripts, Autoconf scripts can support cross-compiling, if some care is taken in writing them.

Autoconf does not solve all problems related to making portable software packages—for a more complete solution, it should be used in concert with other GNU build tools like Automake and Libtool. These other tools take on jobs like the creation of a portable, recursive Makefile with all of the standard targets, linking of shared libraries, and so on. See The GNU Build System, for more information.

Autoconf imposes some restrictions on the names of macros used with #if in C programs (see Preprocessor Symbol Index).

Autoconf requires GNU M4 in order to generate the scripts. It uses features that some unix versions of M4, including GNU M4 1.3, do not have. You must use version 1.4 or later of GNU M4.

See Autoconf 1, for information about upgrading from version 1. See History, for the story of Autoconf's development. See FAQ, for answers to some common questions about Autoconf.

See the Autoconf web page for up-to-date information, details on the mailing lists, pointers to a list of known bugs, etc.

Mail suggestions to the Autoconf mailing list.

Bug reports should be preferably submitted to the Autoconf Gnats database, or sent to the Autoconf Bugs mailing list. If possible, first check that your bug is not already solved in current development versions, and that it has not been reported yet. Be sure to include all the needed information and a short that demonstrates the problem.

Autoconf's development tree is accessible via CVS; see the Autoconf web page for details. There is also a CVSweb interface to the Autoconf development tree. Patches relative to the current CVS version can be sent for review to the Autoconf Patches mailing list.

Because of its mission, Autoconf includes only a set of often-used macros that have already demonstrated their usefulness. Nevertheless, if you wish to share your macros, or find existing ones, see the Autoconf Macro Archive, which is kindly run by Peter Simons.

2 The GNU Build System

Autoconf solves an important problem—reliable discovery of system-specific build and run-time information—but this is only one piece of the puzzle for the development of portable software. To this end, the GNU project has developed a suite of integrated utilities to finish the job Autoconf started: the GNU build system, whose most important components are Autoconf, Automake, and Libtool. In this chapter, we introduce you to those tools, point you to sources of more information, and try to convince you to use the entire GNU build system for your software.

2.1 Automake

The ubiquity of make means that a Makefile is almost the only viable way to distribute automatic build rules for software, but one quickly runs into make's numerous limitations. Its lack of support for automatic dependency tracking, recursive builds in subdirectories, reliable timestamps (e.g., for network filesystems), and so on, mean that developers must painfully (and often incorrectly) reinvent the wheel for each project. Portability is non-trivial, thanks to the quirks of make on many systems. On top of all this is the manual labor required to implement the many standard targets that users have come to expect (make install, make distclean, make uninstall, etc.). Since you are, of course, using Autoconf, you also have to insert repetitive code in your to recognize @CC@, @CFLAGS@, and other substitutions provided by configure. Into this mess steps Automake. Automake allows you to specify your build needs in a file with a vastly simpler and more powerful syntax than that of a plain Makefile, and then generates a portable for use with Autoconf. For example, the to build and install a simple “Hello world” program might look like:

     bin_PROGRAMS = hello
     hello_SOURCES = hello.c

The resulting (~400 lines) automatically supports all the standard targets, the substitutions provided by Autoconf, automatic dependency tracking, VPATH building, and so on. make will build the hello program, and make install will install it in /usr/local/bin (or whatever prefix was given to configure, if not /usr/local).

The benefits of Automake increase for larger packages (especially ones with subdirectories), but even for small programs the added convenience and portability can be substantial. And that's not all....

2.2 Libtool

Very often, one wants to build not only programs, but libraries, so that other programs can benefit from the fruits of your labor. Ideally, one would like to produce shared (dynamically linked) libraries, which can be used by multiple programs without duplication on disk or in memory and can be updated independently of the linked programs. Producing shared libraries portably, however, is the stuff of nightmares—each system has its own incompatible tools, compiler flags, and magic incantations. Fortunately, GNU provides a solution: Libtool. Libtool handles all the requirements of building shared libraries for you, and at this time seems to be the only way to do so with any portability. It also handles many other headaches, such as: the interaction of Makefile rules with the variable suffixes of shared libraries, linking reliably with shared libraries before they are installed by the superuser, and supplying a consistent versioning system (so that different versions of a library can be installed or upgraded without breaking binary compatibility). Although Libtool, like Autoconf, can be used on its own, it is most simply utilized in conjunction with Automake—there, Libtool is used automatically whenever shared libraries are needed, and you need not know its syntax.

2.3 Pointers

Developers who are used to the simplicity of make for small projects on a single system might be daunted at the prospect of learning to use Automake and Autoconf. As your software is distributed to more and more users, however, you will otherwise quickly find yourself putting lots of effort into reinventing the services that the GNU build tools provide, and making the same mistakes that they once made and overcame. (Besides, since you're already learning Autoconf, Automake will be a piece of cake.)

There are a number of places that you can go to for more information on the GNU build tools.

3 Making configure Scripts

The configuration scripts that Autoconf produces are by convention called configure. When run, configure creates several files, replacing configuration parameters in them with appropriate values. The files that configure creates are:

To create a configure script with Autoconf, you need to write an Autoconf input file (or and run autoconf on it. If you write your own feature tests to supplement those that come with Autoconf, you might also write files called aclocal.m4 and acsite.m4. If you use a C header file to contain #define directives, you might also run autoheader, and you will distribute the generated file with the package.

Here is a diagram showing how the files that can be used in configuration are produced. Programs that are executed are suffixed by *. Optional files are enclosed in square brackets ([]). autoconf and autoheader also read the installed Autoconf macro files (by reading autoconf.m4).

Files used in preparing a software package for distribution:

     your source files --> [autoscan*] --> [configure.scan] -->
                    |   .------> autoconf* -----> configure
     [aclocal.m4] --+---+
                    |   `-----> [autoheader*] --> []
     [acsite.m4] ---'

Files used in configuring a software package:

                            .-------------> [config.cache]
     configure* ------------+-------------> config.log
     [] -.       v            .-> [config.h] -.
                    +--> config.status* -+               +--> make* ---'                    `-> Makefile ---'

3.1 Writing

To produce a configure script for a software package, create a file called that contains invocations of the Autoconf macros that test the system features your package needs or can use. Autoconf macros already exist to check for many features; see Existing Tests, for their descriptions. For most other features, you can use Autoconf template macros to produce custom checks; see Writing Tests, for information about them. For especially tricky or specialized features, might need to contain some hand-crafted shell commands; see Portable Shell. The autoscan program can give you a good start in writing (see autoscan Invocation, for more information).

Previous versions of Autoconf promoted the name, which is somewhat ambiguous (the tool needed to process this file is not described by its extension), and introduces a slight confusion with and so on (for which .in means “to be processed by configure”). Using is now preferred.

3.1.1 A Shell Script Compiler

Just as for any other computer language, in order to properly program in Autoconf you must understand what problem the language tries to address and how it does so.

The problem Autoconf addresses is that the world is a mess. After all, you are using Autoconf in order to have your package compile easily on all sorts of different systems, some of them being extremely hostile. Autoconf itself bears the price for these differences: configure must run on all those systems, and thus configure must limit itself to their lowest common denominator of features.

Naturally, you might then think of shell scripts; who needs autoconf? A set of properly written shell functions is enough to make it easy to write configure scripts by hand. Sigh! Unfortunately, shell functions do not belong to the least common denominator; therefore, where you would like to define a function and use it ten times, you would instead need to copy its body ten times.

So, what is really needed is some kind of compiler, autoconf, that takes an Autoconf program,, and transforms it into a portable shell script, configure.

How does autoconf perform this task?

There are two obvious possibilities: creating a brand new language or extending an existing one. The former option is very attractive: all sorts of optimizations could easily be implemented in the compiler and many rigorous checks could be performed on the Autoconf program (e.g., rejecting any non-portable construct). Alternatively, you can extend an existing language, such as the sh (Bourne shell) language.

Autoconf does the latter: it is a layer on top of sh. It was therefore most convenient to implement autoconf as a macro expander: a program that repeatedly performs macro expansions on text input, replacing macro calls with macro bodies and producing a pure sh script in the end. Instead of implementing a dedicated Autoconf macro expander, it is natural to use an existing general-purpose macro language, such as M4, and implement the extensions as a set of M4 macros.

3.1.2 The Autoconf Language

The Autoconf language is very different from many other computer languages because it treats actual code the same as plain text. Whereas in C, for instance, data and instructions have very different syntactic status, in Autoconf their status is rigorously the same. Therefore, we need a means to distinguish literal strings from text to be expanded: quotation.

When calling macros that take arguments, there must not be any blank space between the macro name and the open parenthesis. Arguments should be enclosed within the M4 quote characters [ and ], and be separated by commas. Any leading spaces in arguments are ignored, unless they are quoted. You may safely leave out the quotes when the argument is simple text, but always quote complex arguments such as other macro calls. This rule applies recursively for every macro call, including macros called from other macros.

For instance:

                     [AC_MSG_ERROR([Sorry, can't do anything for you])])

is quoted properly. You may safely simplify its quotation to:

                     [AC_MSG_ERROR([Sorry, can't do anything for you])])

Notice that the argument of AC_MSG_ERROR is still quoted; otherwise, its comma would have been interpreted as an argument separator.

The following example is wrong and dangerous, as it is underquoted:

                     AC_MSG_ERROR([Sorry, can't do anything for you]))

In other cases, you may have to use text that also resembles a macro call. You must quote that text even when it is not passed as a macro argument:

     echo "Hard rock was here!  --[AC_DC]"

which will result in

     echo "Hard rock was here!  --AC_DC"

When you use the same text in a macro argument, you must therefore have an extra quotation level (since one is stripped away by the macro substitution). In general, then, it is a good idea to use double quoting for all literal string arguments:

     AC_MSG_WARN([[AC_DC stinks  --Iron Maiden]])

You are now able to understand one of the constructs of Autoconf that has been continually misunderstood... The rule of thumb is that whenever you expect macro expansion, expect quote expansion; i.e., expect one level of quotes to be lost. For instance:

     AC_COMPILE_IFELSE([char b[10];],, [AC_MSG_ERROR([you lose])])

is incorrect: here, the first argument of AC_COMPILE_IFELSE is char b[10]; and will be expanded once, which results in char b10;. (There was an idiom common in Autoconf's past to address this issue via the M4 changequote primitive, but do not use it!) Let's take a closer look: the author meant the first argument to be understood as a literal, and therefore it must be quoted twice:

     AC_COMPILE_IFELSE([[char b[10];]],, [AC_MSG_ERROR([you lose])])

Voilà, you actually produce char b[10]; this time!

The careful reader will notice that, according to these guidelines, the “properly” quoted AC_CHECK_HEADER example above is actually lacking three pairs of quotes! Nevertheless, for the sake of readability, double quotation of literals is used only where needed in this manual.

Some macros take optional arguments, which this documentation represents as [arg] (not to be confused with the quote characters). You may just leave them empty, or use [] to make the emptiness of the argument explicit, or you may simply omit the trailing commas. The three lines below are equivalent:

     AC_CHECK_HEADERS(stdio.h, [], [], [])

It is best to put each macro call on its own line in Most of the macros don't add extra newlines; they rely on the newline after the macro call to terminate the commands. This approach makes the generated configure script a little easier to read by not inserting lots of blank lines. It is generally safe to set shell variables on the same line as a macro call, because the shell allows assignments without intervening newlines.

You can include comments in files by starting them with the #. For example, it is helpful to begin files with a line like this:

     # Process this file with autoconf to produce a configure script.

3.1.3 Standard Layout

The order in which calls the Autoconf macros is not important, with a few exceptions. Every must contain a call to AC_INIT before the checks, and a call to AC_OUTPUT at the end (see Output). Additionally, some macros rely on other macros having been called first, because they check previously set values of some variables to decide what to do. These macros are noted in the individual descriptions (see Existing Tests), and they also warn you when configure is created if they are called out of order.

To encourage consistency, here is a suggested order for calling the Autoconf macros. Generally speaking, the things near the end of this list are those that could depend on things earlier in it. For example, library functions could be affected by types and libraries.

     Autoconf requirements
     AC_INIT(package, version, bug-report-address)
     information on the package
     checks for programs
     checks for libraries
     checks for header files
     checks for types
     checks for structures
     checks for compiler characteristics
     checks for library functions
     checks for system services

3.2 Using autoscan to Create

The autoscan program can help you create and/or maintain a file for a software package. autoscan examines source files in the directory tree rooted at a directory given as a command line argument, or the current directory if none is given. It searches the source files for common portability problems and creates a file configure.scan which is a preliminary for that package, and checks a possibly existing for completeness.

When using autoscan to create a, you should manually examine configure.scan before renaming it to; it will probably need some adjustments. Occasionally, autoscan outputs a macro in the wrong order relative to another macro, so that autoconf produces a warning; you need to move such macros manually. Also, if you want the package to use a configuration header file, you must add a call to AC_CONFIG_HEADERS (see Configuration Headers). You might also have to change or add some #if directives to your program in order to make it work with Autoconf (see ifnames Invocation, for information about a program that can help with that job).

When using autoscan to maintain a, simply consider adding its suggestions. The file autoscan.log will contain detailed information on why a macro is requested.

autoscan uses several data files (installed along with Autoconf) to determine which macros to output when it finds particular symbols in a package's source files. These data files all have the same format: each line consists of a symbol, whitespace, and the Autoconf macro to output if that symbol is encountered. Lines starting with # are comments.

autoscan accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Print the names of the files it examines and the potentially interesting symbols it finds in them. This output can be voluminous.
-I dir
Append dir to the include path. Multiple invocations accumulate.
-B dir
Prepend dir to the include path. Multiple invocations accumulate.

3.3 Using ifnames to List Conditionals

ifnames can help you write for a software package. It prints the identifiers that the package already uses in C preprocessor conditionals. If a package has already been set up to have some portability, ifnames can thus help you figure out what its configure needs to check for. It may help fill in some gaps in a generated by autoscan (see autoscan Invocation).

ifnames scans all of the C source files named on the command line (or the standard input, if none are given) and writes to the standard output a sorted list of all the identifiers that appear in those files in #if, #elif, #ifdef, or #ifndef directives. It prints each identifier on a line, followed by a space-separated list of the files in which that identifier occurs.

ifnames accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.

3.4 Using autoconf to Create configure

To create configure from, run the autoconf program with no arguments. autoconf processes with the M4 macro processor, using the Autoconf macros. If you give autoconf an argument, it reads that file instead of and writes the configuration script to the standard output instead of to configure. If you give autoconf the argument -, it reads from the standard input instead of and writes the configuration script to the standard output.

The Autoconf macros are defined in several files. Some of the files are distributed with Autoconf; autoconf reads them first. Then it looks for the optional file acsite.m4 in the directory that contains the distributed Autoconf macro files, and for the optional file aclocal.m4 in the current directory. Those files can contain your site's or the package's own Autoconf macro definitions (see Writing Autoconf Macros, for more information). If a macro is defined in more than one of the files that autoconf reads, the last definition it reads overrides the earlier ones.

autoconf accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Report processing steps.
Don't remove the temporary files.
Remake configure even if newer than its input files.
-I dir
Append dir to the include path. Multiple invocations accumulate.
-B dir
Prepend dir to the include path. Multiple invocations accumulate.
-o file
Save output (script or trace) to file. The file - stands for the standard output.
-W category
Report the warnings related to category (which can actually be a comma separated list). See Reporting Messages, macro AC_DIAGNOSE, for a comprehensive list of categories. Special values include:
report all the warnings
report none
treats warnings as errors
disable warnings falling into category

Warnings about syntax are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored. Passing -W category will actually behave as if you had passed --warnings=syntax,$WARNINGS,category. If you want to disable the defaults and WARNINGS, but (for example) enable the warnings about obsolete constructs, you would use -W none,obsolete.

Because autoconf uses autom4te behind the scenes, it displays a back trace for errors, but not for warnings; if you want them, just pass -W error. See autom4te Invocation, for some examples.

-t macro[:format]
Do not create the configure script, but list the calls to macro according to the format. Multiple --trace arguments can be used to list several macros. Multiple --trace arguments for a single macro are not cumulative; instead, you should just make format as long as needed.

The format is a regular string, with newlines if desired, and several special escape codes. It defaults to $f:$l:$n:$%; see autom4te Invocation, for details on the format.

By default, --trace does not trace the initialization of the Autoconf macros (typically the AC_DEFUN definitions). This results in a noticeable speedup, but can be disabled by this option.

It is often necessary to check the content of a file, but parsing it yourself is extremely fragile and error-prone. It is suggested that you rely upon --trace to scan For instance, to find the list of variables that are substituted, use:

     $ autoconf -t AC_SUBST
     More traces deleted

The example below highlights the difference between $@, $*, and $%.

     $ cat
     AC_DEFINE(This, is, [an
     $ autoconf -t 'AC_DEFINE:@: $@
     *: $*
     $: $%'
     @: [This],[is],[an
     *: This,is,an
     $: This:is:an [example]

The format gives you a lot of freedom:

     $ autoconf -t 'AC_SUBST:$$ac_subst{"$1"} = "$f:$l";'
     $ac_subst{"ECHO_C"} = "";
     $ac_subst{"ECHO_N"} = "";
     $ac_subst{"ECHO_T"} = "";
     More traces deleted

A long separator can be used to improve the readability of complex structures, and to ease their parsing (for instance when no single character is suitable as a separator):

     $ autoconf -t 'AM_MISSING_PROG:${|:::::|}*'
     More traces deleted

3.5 Using autoreconf to Update configure Scripts

Installing the various components of the GNU Build System can be tedious: running autopoint for Gettext, automake for etc. in each directory. It may be needed either because some tools such as automake have been updated on your system, or because some of the sources such as have been updated, or finally, simply in order to install the GNU Build System in a fresh tree.

autoreconf runs autoconf, autoheader, aclocal, automake, libtoolize, and autopoint (when appropriate) repeatedly to update the GNU Build System in the specified directories and their subdirectories (see Subdirectories). By default, it only remakes those files that are older than their sources.

If you install a new version of some tool, you can make autoreconf remake all of the files by giving it the --force option.

See Automatic Remaking, for Makefile rules to automatically remake configure scripts when their source files change. That method handles the timestamps of configuration header templates properly, but does not pass --autoconf-dir=dir or --localdir=dir.

autoreconf accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Print the name of each directory where autoreconf runs autoconf (and autoheader, if appropriate).
Don't remove the temporary files.
Remake even configure scripts and configuration headers that are newer than their input files ( and, if present, aclocal.m4).
Install the missing auxiliary files in the package. By default, files are copied; this can be changed with --symlink.

This option triggers calls to automake --add-missing, libtoolize, autopoint, etc.

When used with --install, install symbolic links to the missing auxiliary files instead of copying them.
When the directories were configured, update the configuration by running ./config.status --recheck && ./config.status, and then run make.
-I dir
Append dir to the include path. Multiple invocations accumulate.
-B dir
Prepend dir to the include path. Multiple invocations accumulate.
-W category
Report the warnings related to category (which can actually be a comma separated list).
related to cross compilation issues.
report the uses of obsolete constructs.
portability issues
dubious syntactic constructs.
report all the warnings
report none
treats warnings as errors
disable warnings falling into category

Warnings about syntax are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored. Passing -W category will actually behave as if you had passed --warnings=syntax,$WARNINGS,category. If you want to disable the defaults and WARNINGS, but (for example) enable the warnings about obsolete constructs, you would use -W none,obsolete.

4 Initialization and Output Files

Autoconf-generated configure scripts need some information about how to initialize, such as how to find the package's source files and about the output files to produce. The following sections describe the initialization and the creation of output files.

4.1 Initializing configure

Every configure script must call AC_INIT before doing anything else. The only other required macro is AC_OUTPUT (see Output).

— Macro: AC_INIT (package, version, [bug-report], [tarname])

Process any command-line arguments and perform various initializations and verifications.

Set the name of the package and its version. These are typically used in --version support, including that of configure. The optional argument bug-report should be the email to which users should send bug reports. The package tarname differs from package: the latter designates the full package name (e.g., GNU Autoconf), while the former is meant for distribution tar ball names (e.g., autoconf). It defaults to package with GNU stripped, lower-cased, and all characters other than alphanumerics and underscores are changed to -.

It is preferable that the arguments of AC_INIT be static, i.e., there should not be any shell computation, but they can be computed by M4.

The following M4 macros (e.g., AC_PACKAGE_NAME), output variables (e.g., PACKAGE_NAME), and preprocessor symbols (e.g., PACKAGE_NAME) are defined by AC_INIT:

Exactly package.
Exactly tarname.
Exactly version.
Exactly package version.
Exactly bug-report.

4.2 Notices in configure

The following macros manage version numbers for configure scripts. Using them is optional.

— Macro: AC_PREREQ (version)

Ensure that a recent enough version of Autoconf is being used. If the version of Autoconf being used to create configure is earlier than version, print an error message to the standard error output and exit with failure (exit status is 63). For example:


This macro is the only macro that may be used before AC_INIT, but for consistency, you are invited not to do so.

— Macro: AC_COPYRIGHT (copyright-notice)

State that, in addition to the Free Software Foundation's copyright on the Autoconf macros, parts of your configure are covered by the copyright-notice.

The copyright-notice will show up in both the head of configure and in configure --version.

— Macro: AC_REVISION (revision-info)

Copy revision stamp revision-info into the configure script, with any dollar signs or double-quotes removed. This macro lets you put a revision stamp from into configure without RCS or CVS changing it when you check in configure. That way, you can determine easily which revision of a particular configure corresponds to.

For example, this line in

          AC_REVISION($Revision: 1.30 $)

produces this in configure:

          #! /bin/sh
          # From Revision: 1.30

4.3 Finding configure Input

— Macro: AC_CONFIG_SRCDIR (unique-file-in-source-dir)

unique-file-in-source-dir is some file that is in the package's source directory; configure checks for this file's existence to make sure that the directory that it is told contains the source code in fact does. Occasionally people accidentally specify the wrong directory with --srcdir; this is a safety check. See configure Invocation, for more information.

Packages that do manual configuration or use the install program might need to tell configure where to find some other shell scripts by calling AC_CONFIG_AUX_DIR, though the default places it looks are correct for most cases.

— Macro: AC_CONFIG_AUX_DIR (dir)

Use the auxiliary build tools (e.g., install-sh, config.sub, config.guess, Cygnus configure, Automake and Libtool scripts etc.) that are in directory dir. These are auxiliary files used in configuration. dir can be either absolute or relative to srcdir. The default is srcdir or srcdir/.. or srcdir/../.., whichever is the first that contains install-sh. The other files are not checked for, so that using AC_PROG_INSTALL does not automatically require distributing the other auxiliary files. It checks for also, but that name is obsolete because some make have a rule that creates install from it if there is no Makefile.

The auxiliary directory should not be named aux for portability to MS-DOS, because the filename aux is reserved under MS-DOS.

Similarly, packages that use aclocal should declare where local macros can be found using AC_CONFIG_MACRO_DIR.

— Macro: AC_CONFIG_MACRO_DIR (dir)

Future versions of autopoint, libtoolize, aclocal and autoreconf will use directory dir as the location of additional local Autoconf macros. Be sure to call this macro directly from so that tools that install macros for aclocal can find the declaration before --trace can be called safely.

4.4 Outputting Files

Every Autoconf script, e.g.,, should finish by calling AC_OUTPUT. That is the macro that generates and runs config.status, which will create the Makefiles and any other files resulting from configuration. This is the only required macro besides AC_INIT (see Input).

— Macro: AC_OUTPUT

Generate config.status and launch it. Call this macro once, at the end of

config.status will perform all the configuration actions: all the output files (see Configuration Files, macro AC_CONFIG_FILES), header files (see Configuration Headers, macro AC_CONFIG_HEADERS), commands (see Configuration Commands, macro AC_CONFIG_COMMANDS), links (see Configuration Links, macro AC_CONFIG_LINKS), subdirectories to configure (see Subdirectories, macro AC_CONFIG_SUBDIRS) are honored.

The location of your AC_OUTPUT invocation is the exact point where configuration actions are taken: any code afterwards will be executed by configure once config.status was run. If you want to bind actions to config.status itself (independently of whether configure is being run), see Running Arbitrary Configuration Commands.

Historically, the usage of AC_OUTPUT was somewhat different. See Obsolete Macros, for a description of the arguments that AC_OUTPUT used to support.

If you run make in subdirectories, you should run it using the make variable MAKE. Most versions of make set MAKE to the name of the make program plus any options it was given. (But many do not include in it the values of any variables set on the command line, so those are not passed on automatically.) Some old versions of make do not set this variable. The following macro allows you to use it even with those versions.


If make predefines the Make variable MAKE, define output variable SET_MAKE to be empty. Otherwise, define SET_MAKE to contain MAKE=make. Calls AC_SUBST for SET_MAKE.

If you use this macro, place a line like this in each that runs MAKE on other directories:


4.5 Performing Configuration Actions

configure is designed so that it appears to do everything itself, but there is actually a hidden slave: config.status. configure is in charge of examining your system, but it is config.status that actually takes the proper actions based on the results of configure. The most typical task of config.status is to instantiate files.

This section describes the common behavior of the four standard instantiating macros: AC_CONFIG_FILES, AC_CONFIG_HEADERS, AC_CONFIG_COMMANDS and AC_CONFIG_LINKS. They all have this prototype:

     AC_CONFIG_FOOS(tag..., [commands], [init-cmds])

where the arguments are:

A whitespace-separated list of tags, which are typically the names of the files to instantiate.

You are encouraged to use literals as tags. In particular, you should avoid

          ... && my_foos="$my_foos fooo"
          ... && my_foos="$my_foos foooo"

and use this instead:

          ... && AC_CONFIG_FOOS(fooo)
          ... && AC_CONFIG_FOOS(foooo)

The macros AC_CONFIG_FILES and AC_CONFIG_HEADERS use special tags: they may have the form output or output:inputs. The file output is instantiated from its templates, inputs (defaulting to

For instance AC_CONFIG_FILES(Makefile:boiler/ asks for the creation of Makefile that will be the expansion of the output variables in the concatenation of boiler/ and boiler/

The special value - might be used to denote the standard output when used in output, or the standard input when used in the inputs. You most probably don't need to use this in, but it is convenient when using the command line interface of ./config.status, see config.status Invocation, for more details.

The inputs may be absolute or relative filenames. In the latter case they are first looked for in the build tree, and then in the source tree.

Shell commands output literally into config.status, and associated with a tag that the user can use to tell config.status which the commands to run. The commands are run each time a tag request is given to config.status, typically each time the file tag is created.

The variables set during the execution of configure are not available here: you first need to set them via the init-cmds. Nonetheless the following variables are precomputed:

The path from the top build directory to the top source directory. This is what configure's option --srcdir sets.
The path from the current build directory to the top source directory.
The path from the current build directory to the top build directory. It can be empty, or else ends with a slash, so that you may concatenate it.
The path from the current build directory to the corresponding source directory.

The current directory refers to the directory (or pseudo-directory) containing the input part of tags. For instance, running

          AC_CONFIG_COMMANDS([deep/dir/out:in/], [...], [...])

with --srcdir=../package produces the following values:

          # Argument of --srcdir
          # Reversing deep/dir
          # Concatenation of $ac_top_builddir and srcdir
          # Concatenation of $ac_top_srcdir and deep/dir

independently of in/

Shell commands output unquoted near the beginning of config.status, and executed each time config.status runs (regardless of the tag). Because they are unquoted, for example, $var will be output as the value of var. init-cmds is typically used by configure to give config.status some variables it needs to run the commands.

You should be extremely cautious in your variable names: all the init-cmds share the same name space and may overwrite each other in unpredictable ways. Sorry....

All these macros can be called multiple times, with different tags, of course!

4.6 Creating Configuration Files

Be sure to read the previous section, Configuration Actions.

— Macro: AC_CONFIG_FILES (file..., [cmds], [init-cmds])

Make AC_OUTPUT create each file by copying an input file (by default, substituting the output variable values. This macro is one of the instantiating macros; see Configuration Actions. See Makefile Substitutions, for more information on using output variables. See Setting Output Variables, for more information on creating them. This macro creates the directory that the file is in if it doesn't exist. Usually, Makefiles are created this way, but other files, such as .gdbinit, can be specified as well.

Typical calls to AC_CONFIG_FILES look like this:

          AC_CONFIG_FILES([Makefile src/Makefile man/Makefile X/Imakefile])
          AC_CONFIG_FILES([autoconf], [chmod +x autoconf])

You can override an input file name by appending to file a colon-separated list of input files. Examples:


Doing this allows you to keep your file names acceptable to MS-DOS, or to prepend and/or append boilerplate to the file.

4.7 Substitutions in Makefiles

Each subdirectory in a distribution that contains something to be compiled or installed should come with a file, from which configure will create a Makefile in that directory. To create a Makefile, configure performs a simple variable substitution, replacing occurrences of @variable@ in with the value that configure has determined for that variable. Variables that are substituted into output files in this way are called output variables. They are ordinary shell variables that are set in configure. To make configure substitute a particular variable into the output files, the macro AC_SUBST must be called with that variable name as an argument. Any occurrences of @variable@ for other variables are left unchanged. See Setting Output Variables, for more information on creating output variables with AC_SUBST.

A software package that uses a configure script should be distributed with a file, but no Makefile; that way, the user has to properly configure the package for the local system before compiling it.

See Makefile Conventions (The GNU Coding Standards), for more information on what to put in Makefiles.

4.7.1 Preset Output Variables

Some output variables are preset by the Autoconf macros. Some of the Autoconf macros set additional output variables, which are mentioned in the descriptions for those macros. See Output Variable Index, for a complete list of output variables. See Installation Directory Variables, for the list of the preset ones related to installation directories. Below are listed the other preset ones. They all are precious variables (see Setting Output Variables, AC_ARG_VAR).

— Variable: CFLAGS

Debugging and optimization options for the C compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_CC (or empty if you don't). configure uses this variable when compiling programs to test for C features.

— Variable: configure_input

A comment saying that the file was generated automatically by configure and giving the name of the input file. AC_OUTPUT adds a comment line containing this variable to the top of every Makefile it creates. For other files, you should reference this variable in a comment at the top of each input file. For example, an input shell script should begin like this:

          #! /bin/sh
          # @configure_input@

The presence of that line also reminds people editing the file that it needs to be processed by configure in order to be used.

— Variable: CPPFLAGS

Header file search directory (-Idir) and any other miscellaneous options for the C and C++ preprocessors and compilers. If it is not set in the environment when configure runs, the default value is empty. configure uses this variable when compiling or preprocessing programs to test for C and C++ features.

— Variable: CXXFLAGS

Debugging and optimization options for the C++ compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_CXX (or empty if you don't). configure uses this variable when compiling programs to test for C++ features.

— Variable: DEFS

-D options to pass to the C compiler. If AC_CONFIG_HEADERS is called, configure replaces @DEFS@ with -DHAVE_CONFIG_H instead (see Configuration Headers). This variable is not defined while configure is performing its tests, only when creating the output files. See Setting Output Variables, for how to check the results of previous tests.

— Variable: ECHO_C
— Variable: ECHO_N
— Variable: ECHO_T

How does one suppress the trailing newline from echo for question-answer message pairs? These variables provide a way:

          echo $ECHO_N "And the winner is... $ECHO_C"
          sleep 100000000000
          echo "${ECHO_T}dead."

Some old and uncommon echo implementations offer no means to achieve this, in which case ECHO_T is set to tab. You might not want to use it.

— Variable: FCFLAGS

Debugging and optimization options for the Fortran compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_FC (or empty if you don't). configure uses this variable when compiling programs to test for Fortran features.

— Variable: FFLAGS

Debugging and optimization options for the Fortran 77 compiler. If it is not set in the environment when configure runs, the default value is set when you call AC_PROG_F77 (or empty if you don't). configure uses this variable when compiling programs to test for Fortran 77 features.

— Variable: LDFLAGS

Stripping (-s), path (-L), and any other miscellaneous options for the linker. Don't use this variable to pass library names (-l) to the linker, use LIBS instead. If it is not set in the environment when configure runs, the default value is empty. configure uses this variable when linking programs to test for C, C++, and Fortran features.

— Variable: LIBS

-l options to pass to the linker. The default value is empty, but some Autoconf macros may prepend extra libraries to this variable if those libraries are found and provide necessary functions, see Libraries. configure uses this variable when linking programs to test for C, C++, and Fortran features.

— Variable: builddir

Rigorously equal to .. Added for symmetry only.

— Variable: abs_builddir

Absolute path of builddir.

— Variable: top_builddir

The relative path to the top-level of the current build tree. In the top-level directory, this is the same as builddir.

— Variable: abs_top_builddir

Absolute path of top_builddir.

— Variable: srcdir

The relative path to the directory that contains the source code for that Makefile.

— Variable: abs_srcdir

Absolute path of srcdir.

— Variable: top_srcdir

The relative path to the top-level source code directory for the package. In the top-level directory, this is the same as srcdir.

— Variable: abs_top_srcdir

Absolute path of top_srcdir.

4.7.2 Installation Directory Variables

The following variables specify the directories where the package will be installed, see Variables for Installation Directories (The GNU Coding Standards), for more information. See the end of this section for details on when and how to use these variables.

— Variable: bindir

The directory for installing executables that users run.

— Variable: datadir

The directory for installing read-only architecture-independent data.

— Variable: exec_prefix

The installation prefix for architecture-dependent files. By default it's the same as prefix. You should avoid installing anything directly to exec_prefix. However, the default value for directories containing architecture-dependent files should be relative to exec_prefix.

— Variable: includedir

The directory for installing C header files.

— Variable: infodir

The directory for installing documentation in Info format.

— Variable: libdir

The directory for installing object code libraries.

— Variable: libexecdir

The directory for installing executables that other programs run.

— Variable: localstatedir

The directory for installing modifiable single-machine data.

— Variable: mandir

The top-level directory for installing documentation in man format.

— Variable: oldincludedir

The directory for installing C header files for non-GCC compilers.

— Variable: prefix

The common installation prefix for all files. If exec_prefix is defined to a different value, prefix is used only for architecture-independent files.

— Variable: sbindir

The directory for installing executables that system administrators run.

— Variable: sharedstatedir

The directory for installing modifiable architecture-independent data.

— Variable: sysconfdir

The directory for installing read-only single-machine data.

Most of these variables have values that rely on prefix or exec_prefix. It is deliberate that the directory output variables keep them unexpanded: typically @datadir@ will be replaced by ${prefix}/share, not /usr/local/share.

This behavior is mandated by the GNU coding standards, so that when the user runs:

she can still specify a different prefix from the one specified to configure, in which case, if needed, the package shall hard code dependencies corresponding to the make-specified prefix.
make install
she can specify a different installation location, in which case the package must still depend on the location which was compiled in (i.e., never recompile when make install is run). This is an extremely important feature, as many people may decide to install all the files of a package grouped together, and then install links from the final locations to there.

In order to support these features, it is essential that datadir remains being defined as ${prefix}/share to depend upon the current value of prefix.

A corollary is that you should not use these variables except in Makefiles. For instance, instead of trying to evaluate datadir in configure and hard-coding it in Makefiles using e.g., AC_DEFINE_UNQUOTED(DATADIR, "$datadir"), you should add -DDATADIR="$(datadir)" to your CPPFLAGS.

Similarly you should not rely on AC_OUTPUT_FILES to replace datadir and friends in your shell scripts and other files, rather let make manage their replacement. For instance Autoconf ships templates of its shell scripts ending with .in, and uses a Makefile snippet similar to:

     edit = sed \
             -e 's,@datadir\@,$(pkgdatadir),g' \
             -e 's,@prefix\@,$(prefix),g'
     autoconf: Makefile $(srcdir)/
             rm -f autoconf autoconf.tmp
             $(edit) $(srcdir)/ >autoconf.tmp
             chmod +x autoconf.tmp
             mv autoconf.tmp autoconf
     autoheader: Makefile $(srcdir)/
             rm -f autoheader autoheader.tmp
             $(edit) $(srcdir)/ >autoheader.tmp
             chmod +x autoheader.tmp
             mv autoheader.tmp autoheader

Some details are noteworthy:

The backslash prevents configure from replacing @datadir@ in the sed expression itself.
Don't use @pkgdatadir@! Use the matching makefile variable instead.
Don't use / in the sed expression(s) since most likely the variables you use, such as $(pkgdatadir), will contain some.
Dependency on Makefile
Since edit uses values that depend on the configuration specific values (prefix etc.) and not only on VERSION and so forth, the output depends on Makefile, not
Separated dependencies and Single Suffix Rules
You can't use them! The above snippet cannot be (portably) rewritten as:
          autoconf autoheader: Makefile
                  rm -f $@ $@.tmp
                  $(edit) $< >$@.tmp
                  chmod +x $@.tmp
                  mv $@.tmp $@

See Limitations of Make, for details.

Be sure to specify the path to the sources, otherwise the package won't support separated builds.

4.7.3 Build Directories

You can support compiling a software package for several architectures simultaneously from the same copy of the source code. The object files for each architecture are kept in their own directory.

To support doing this, make uses the VPATH variable to find the files that are in the source directory. GNU Make and most other recent make programs can do this. Older make programs do not support VPATH; when using them, the source code must be in the same directory as the object files.

To support VPATH, each should contain two lines that look like:

     srcdir = @srcdir@
     VPATH = @srcdir@

Do not set VPATH to the value of another variable, for example VPATH = $(srcdir), because some versions of make do not do variable substitutions on the value of VPATH.

configure substitutes the correct value for srcdir when it produces Makefile.

Do not use the make variable $<, which expands to the file name of the file in the source directory (found with VPATH), except in implicit rules. (An implicit rule is one such as .c.o, which tells how to create a .o file from a .c file.) Some versions of make do not set $< in explicit rules; they expand it to an empty value.

Instead, Makefile command lines should always refer to source files by prefixing them with $(srcdir)/. For example: time.texinfo
             $(MAKEINFO) $(srcdir)/time.texinfo

4.7.4 Automatic Remaking

You can put rules like the following in the top-level for a package to automatically update the configuration information when you change the configuration files. This example includes all of the optional files, such as aclocal.m4 and those related to configuration header files. Omit from the rules for any of these files that your package does not use.

The $(srcdir)/ prefix is included because of limitations in the VPATH mechanism.

The stamp- files are necessary because the timestamps of and config.h will not be changed if remaking them does not change their contents. This feature avoids unnecessary recompilation. You should include the file your package's distribution, so make will consider up to date. Don't use touch (see Limitations of Usual Tools), rather use echo (using date would cause needless differences, hence CVS conflicts etc.).

     $(srcdir)/configure: aclocal.m4
             cd $(srcdir) && autoconf
     # autoheader might not change, so touch a stamp file.
     $(srcdir)/ aclocal.m4
             cd $(srcdir) && autoheader
             echo timestamp > $(srcdir)/
     config.h: stamp-h
     stamp-h: config.status
     Makefile: config.status
     config.status: configure
             ./config.status --recheck

(Be careful if you copy these lines directly into your Makefile, as you will need to convert the indented lines to start with the tab character.)

In addition, you should use AC_CONFIG_FILES([stamp-h], [echo timestamp > stamp-h]) so config.status will ensure that config.h is considered up to date. See Output, for more information about AC_OUTPUT.

See config.status Invocation, for more examples of handling configuration-related dependencies.

4.8 Configuration Header Files

When a package contains more than a few tests that define C preprocessor symbols, the command lines to pass -D options to the compiler can get quite long. This causes two problems. One is that the make output is hard to visually scan for errors. More seriously, the command lines can exceed the length limits of some operating systems. As an alternative to passing -D options to the compiler, configure scripts can create a C header file containing #define directives. The AC_CONFIG_HEADERS macro selects this kind of output. It should be called right after AC_INIT.

The package should #include the configuration header file before any other header files, to prevent inconsistencies in declarations (for example, if it redefines const). Use #include <config.h> instead of #include "config.h", and pass the C compiler a -I. option (or -I..; whichever directory contains config.h). That way, even if the source directory is configured itself (perhaps to make a distribution), other build directories can also be configured without finding the config.h from the source directory.

— Macro: AC_CONFIG_HEADERS (header ..., [cmds], [init-cmds])

This macro is one of the instantiating macros; see Configuration Actions. Make AC_OUTPUT create the file(s) in the whitespace-separated list header containing C preprocessor #define statements, and replace @DEFS@ in generated files with -DHAVE_CONFIG_H instead of the value of DEFS. The usual name for header is config.h.

If header already exists and its contents are identical to what AC_OUTPUT would put in it, it is left alone. Doing this allows making some changes in the configuration without needlessly causing object files that depend on the header file to be recompiled.

Usually the input file is named; however, you can override the input file name by appending to header a colon-separated list of input files. Examples:


Doing this allows you to keep your file names acceptable to MS-DOS, or to prepend and/or append boilerplate to the file.

See Configuration Actions, for more details on header.

4.8.1 Configuration Header Templates

Your distribution should contain a template file that looks as you want the final header file to look, including comments, with #undef statements which are used as hooks. For example, suppose your makes these calls:


Then you could have code like the following in On systems that have unistd.h, configure will #define HAVE_UNISTD_H to 1. On other systems, the whole line will be commented out (in case the system predefines that symbol).

     /* Define as 1 if you have unistd.h.  */
     #undef HAVE_UNISTD_H

Pay attention that #undef is in the first column, and there is nothing behind HAVE_UNISTD_H, not even white spaces. You can then decode the configuration header using the preprocessor directives:

     #include <conf.h>
     #if HAVE_UNISTD_H
     # include <unistd.h>
     /* We are in trouble.  */

The use of old form templates, with #define instead of #undef is strongly discouraged. Similarly with old templates with comments on the same line as the #undef. Anyway, putting comments in preprocessor macros has never been a good idea.

Since it is a tedious task to keep a template header up to date, you may use autoheader to generate it, see autoheader Invocation.

4.8.2 Using autoheader to Create

The autoheader program can create a template file of C #define statements for configure to use. If invokes AC_CONFIG_HEADERS(file), autoheader creates; if multiple file arguments are given, the first one is used. Otherwise, autoheader creates

In order to do its job, autoheader needs you to document all of the symbols that you might use; i.e., there must be at least one AC_DEFINE or one AC_DEFINE_UNQUOTED call with a third argument for each symbol (see Defining Symbols). An additional constraint is that the first argument of AC_DEFINE must be a literal. Note that all symbols defined by Autoconf's builtin tests are already documented properly; you only need to document those that you define yourself.

You might wonder why autoheader is needed: after all, why would configure need to “patch” a to produce a config.h instead of just creating config.h from scratch? Well, when everything rocks, the answer is just that we are wasting our time maintaining autoheader: generating config.h directly is all that is needed. When things go wrong, however, you'll be thankful for the existence of autoheader.

The fact that the symbols are documented is important in order to check that config.h makes sense. The fact that there is a well-defined list of symbols that should be #define'd (or not) is also important for people who are porting packages to environments where configure cannot be run: they just have to fill in the blanks.

But let's come back to the point: autoheader's invocation...

If you give autoheader an argument, it uses that file instead of and writes the header file to the standard output instead of to If you give autoheader an argument of -, it reads the standard input instead of and writes the header file to the standard output.

autoheader accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Report processing steps.
Don't remove the temporary files.
Remake the template file even if newer than its input files.
-I dir
Append dir to include path. Multiple invocations accumulate.
-B dir
Prepend dir to include path. Multiple invocations accumulate.
-W category
Report the warnings related to category (which can actually be a comma separated list). Current categories include:
report the uses of obsolete constructs
report all the warnings
report none
treats warnings as errors
disable warnings falling into category

4.8.3 Autoheader Macros

autoheader scans and figures out which C preprocessor symbols it might define. It knows how to generate templates for symbols defined by AC_CHECK_HEADERS, AC_CHECK_FUNCS etc., but if you AC_DEFINE any additional symbol, you must define a template for it. If there are missing templates, autoheader fails with an error message.

The simplest way to create a template for a symbol is to supply the description argument to an AC_DEFINE(symbol); see Defining Symbols. You may also use one of the following macros.

— Macro: AH_VERBATIM (key, template)

Tell autoheader to include the template as-is in the header template file. This template is associated with the key, which is used to sort all the different templates and guarantee their uniqueness. It should be a symbol that can be AC_DEFINE'd.

For example:

          [/* Enable GNU extensions on systems that have them.  */
          #ifndef _GNU_SOURCE
          # define _GNU_SOURCE
— Macro: AH_TEMPLATE (key, description)

Tell autoheader to generate a template for key. This macro generates standard templates just like AC_DEFINE when a description is given.

For example:

                      [Define to one of _getb67, GETB67, getb67
                       for Cray-2 and Cray-YMP systems.  This
                       function is required for alloca.c support
                       on those systems.])

will generate the following template, with the description properly justified.

          /* Define to one of _getb67, GETB67, getb67 for Cray-2 and
             Cray-YMP systems.  This function is required for alloca.c
             support on those systems.  */
          #undef CRAY_STACKSEG_END
— Macro: AH_TOP (text)

Include text at the top of the header template file.

— Macro: AH_BOTTOM (text)

Include text at the bottom of the header template file.

4.9 Running Arbitrary Configuration Commands

You can execute arbitrary commands before, during, and after config.status is run. The three following macros accumulate the commands to run when they are called multiple times. AC_CONFIG_COMMANDS replaces the obsolete macro AC_OUTPUT_COMMANDS; see Obsolete Macros, for details.

— Macro: AC_CONFIG_COMMANDS (tag..., [cmds], [init-cmds])

Specify additional shell commands to run at the end of config.status, and shell commands to initialize any variables from configure. Associate the commands with tag. Since typically the cmds create a file, tag should naturally be the name of that file. If needed, the directory hosting tag is created. This macro is one of the instantiating macros; see Configuration Actions.

Here is an unrealistic example:

                             [echo this is extra $fubar, and so on.],

Here is a better one:

          AC_CONFIG_COMMANDS([time-stamp], [date >time-stamp])

Execute the cmds right before creating config.status.


Execute the cmds right after creating config.status.

4.10 Creating Configuration Links

You may find it convenient to create links whose destinations depend upon results of tests. One can use AC_CONFIG_COMMANDS but the creation of relative symbolic links can be delicate when the package is built in a directory different from the source directory.

— Macro: AC_CONFIG_LINKS (dest:source..., [cmds], [init-cmds])

Make AC_OUTPUT link each of the existing files source to the corresponding link name dest. Makes a symbolic link if possible, otherwise a hard link if possible, otherwise a copy. The dest and source names should be relative to the top level source or build directory. This macro is one of the instantiating macros; see Configuration Actions.

For example, this call:


creates in the current directory host.h as a link to srcdir/config/$machine.h, and object.h as a link to srcdir/config/$obj_format.h.

The tempting value . for dest is invalid: it makes it impossible for config.status to guess the links to establish.

One can then run:

          ./config.status host.h object.h

to create the links.

4.11 Configuring Other Packages in Subdirectories

In most situations, calling AC_OUTPUT is sufficient to produce Makefiles in subdirectories. However, configure scripts that control more than one independent package can use AC_CONFIG_SUBDIRS to run configure scripts for other packages in subdirectories.

— Macro: AC_CONFIG_SUBDIRS (dir ...)

Make AC_OUTPUT run configure in each subdirectory dir in the given whitespace-separated list. Each dir should be a literal, i.e., please do not use:

          if test "$package_foo_enabled" = yes; then
            $my_subdirs="$my_subdirs foo"

because this prevents ./configure --help=recursive from displaying the options of the package foo. Rather, you should write:

          if test "$package_foo_enabled" = yes; then

If a given dir is not found, an error is reported: if the subdirectory is optional, write:

          if test -d $srcdir/foo; then

If a given dir contains configure.gnu, it is run instead of configure. This is for packages that might use a non-Autoconf script Configure, which can't be called through a wrapper configure since it would be the same file on case-insensitive filesystems. Likewise, if a dir contains but no configure, the Cygnus configure script found by AC_CONFIG_AUX_DIR is used.

The subdirectory configure scripts are given the same command line options that were given to this configure script, with minor changes if needed, which include:

This macro also sets the output variable subdirs to the list of directories dir .... Makefile rules can use this variable to determine which subdirectories to recurse into.

This macro may be called multiple times.

4.12 Default Prefix

By default, configure sets the prefix for files it installs to /usr/local. The user of configure can select a different prefix using the --prefix and --exec-prefix options. There are two ways to change the default: when creating configure, and when running it.

Some software packages might want to install in a directory other than /usr/local by default. To accomplish that, use the AC_PREFIX_DEFAULT macro.

— Macro: AC_PREFIX_DEFAULT (prefix)

Set the default installation prefix to prefix instead of /usr/local.

It may be convenient for users to have configure guess the installation prefix from the location of a related program that they have already installed. If you wish to do that, you can call AC_PREFIX_PROGRAM.

— Macro: AC_PREFIX_PROGRAM (program)

If the user did not specify an installation prefix (using the --prefix option), guess a value for it by looking for program in PATH, the way the shell does. If program is found, set the prefix to the parent of the directory containing program, else default the prefix as described above (/usr/local or AC_PREFIX_DEFAULT). For example, if program is gcc and the PATH contains /usr/local/gnu/bin/gcc, set the prefix to /usr/local/gnu.

5 Existing Tests

These macros test for particular system features that packages might need or want to use. If you need to test for a kind of feature that none of these macros check for, you can probably do it by calling primitive test macros with appropriate arguments (see Writing Tests).

These tests print messages telling the user which feature they're checking for, and what they find. They cache their results for future configure runs (see Caching Results).

Some of these macros set output variables. See Makefile Substitutions, for how to get their values. The phrase “define name” is used below as a shorthand to mean “define C preprocessor symbol name to the value 1”. See Defining Symbols, for how to get those symbol definitions into your program.

5.1 Common Behavior

Much effort has been expended to make Autoconf easy to learn. The most obvious way to reach this goal is simply to enforce standard interfaces and behaviors, avoiding exceptions as much as possible. Because of history and inertia, unfortunately, there are still too many exceptions in Autoconf; nevertheless, this section describes some of the common rules.

5.1.1 Standard Symbols

All the generic macros that AC_DEFINE a symbol as a result of their test transform their arguments to a standard alphabet. First, argument is converted to upper case and any asterisks (*) are each converted to P. Any remaining characters that are not alphanumeric are converted to underscores.

For instance,

     AC_CHECK_TYPES(struct $Expensive*)

will define the symbol HAVE_STRUCT__EXPENSIVEP if the check succeeds.

5.1.2 Default Includes

Several tests depend upon a set of header files. Since these headers are not universally available, tests actually have to provide a set of protected includes, such as:

     # include <sys/time.h>
     # include <time.h>
     # if HAVE_SYS_TIME_H
     #  include <sys/time.h>
     # else
     #  include <time.h>
     # endif

Unless you know exactly what you are doing, you should avoid using unconditional includes, and check the existence of the headers you include beforehand (see Header Files).

Most generic macros use the following macro to provide the default set of includes:

— Macro: AC_INCLUDES_DEFAULT ([include-directives])

Expand to include-directives if defined, otherwise to:

          #include <stdio.h>
          #if HAVE_SYS_TYPES_H
          # include <sys/types.h>
          #if HAVE_SYS_STAT_H
          # include <sys/stat.h>
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #if HAVE_STRING_H
          # if !STDC_HEADERS && HAVE_MEMORY_H
          #  include <memory.h>
          # endif
          # include <string.h>
          #if HAVE_STRINGS_H
          # include <strings.h>
          #if HAVE_INTTYPES_H
          # include <inttypes.h>
          # if HAVE_STDINT_H
          #  include <stdint.h>
          # endif
          #if HAVE_UNISTD_H
          # include <unistd.h>

If the default includes are used, then check for the presence of these headers and their compatibility, i.e., you don't need to run AC_HEADERS_STDC, nor check for stdlib.h etc.

These headers are checked for in the same order as they are included. For instance, on some systems string.h and strings.h both exist, but conflict. Then HAVE_STRING_H will be defined, but HAVE_STRINGS_H won't.

5.2 Alternative Programs

These macros check for the presence or behavior of particular programs. They are used to choose between several alternative programs and to decide what to do once one has been chosen. If there is no macro specifically defined to check for a program you need, and you don't need to check for any special properties of it, then you can use one of the general program-check macros.

5.2.1 Particular Program Checks

These macros check for particular programs—whether they exist, and in some cases whether they support certain features.

— Macro: AC_PROG_AWK

Check for gawk, mawk, nawk, and awk, in that order, and set output variable AWK to the first one that is found. It tries gawk first because that is reported to be the best implementation.


Check for grep -E and egrep, in that order, and set output variable EGREP to the first one that is found.


Check for grep -F and fgrep, in that order, and set output variable FGREP to the first one that is found.


Set output variable INSTALL to the path of a BSD-compatible install program, if one is found in the current PATH. Otherwise, set INSTALL to dir/install-sh -c, checking the directories specified to AC_CONFIG_AUX_DIR (or its default directories) to determine dir (see Output). Also set the variables INSTALL_PROGRAM and INSTALL_SCRIPT to ${INSTALL} and INSTALL_DATA to ${INSTALL} -m 644.

This macro screens out various instances of install known not to work. It prefers to find a C program rather than a shell script, for speed. Instead of install-sh, it can also use, but that name is obsolete because some make programs have a rule that creates install from it if there is no Makefile.

Autoconf comes with a copy of install-sh that you can use. If you use AC_PROG_INSTALL, you must include either install-sh or in your distribution, or configure will produce an error message saying it can't find them—even if the system you're on has a good install program. This check is a safety measure to prevent you from accidentally leaving that file out, which would prevent your package from installing on systems that don't have a BSD-compatible install program.

If you need to use your own installation program because it has features not found in standard install programs, there is no reason to use AC_PROG_INSTALL; just put the file name of your program into your files.

— Macro: AC_PROG_LEX

If flex is found, set output variable LEX to flex and LEXLIB to -lfl, if that library is in a standard place. Otherwise set LEX to lex and LEXLIB to -ll.

Define YYTEXT_POINTER if yytext is a char * instead of a char []. Also set output variable LEX_OUTPUT_ROOT to the base of the file name that the lexer generates; usually lex.yy, but sometimes something else. These results vary according to whether lex or flex is being used.

You are encouraged to use Flex in your sources, since it is both more pleasant to use than plain Lex and the C source it produces is portable. In order to ensure portability, however, you must either provide a function yywrap or, if you don't use it (e.g., your scanner has no #include-like feature), simply include a %noyywrap statement in the scanner's source. Once this done, the scanner is portable (unless you felt free to use nonportable constructs) and does not depend on any library. In this case, and in this case only, it is suggested that you use this Autoconf snippet:

          if test "$LEX" != flex; then
            LEX="$SHELL $missing_dir/missing flex"
            AC_SUBST(LEX_OUTPUT_ROOT, lex.yy)
            AC_SUBST(LEXLIB, '')

The shell script missing can be found in the Automake distribution.

To ensure backward compatibility, Automake's AM_PROG_LEX invokes (indirectly) this macro twice, which will cause an annoying but benign “AC_PROG_LEX invoked multiple times” warning. Future versions of Automake will fix this issue; meanwhile, just ignore this message.

As part of running the test, this macro may delete any file in the configuration directory named lex.yy.c or lexyy.c.

— Macro: AC_PROG_LN_S

If ln -s works on the current file system (the operating system and file system support symbolic links), set the output variable LN_S to ln -s; otherwise, if ln works, set LN_S to ln, and otherwise set it to cp -p.

If you make a link in a directory other than the current directory, its meaning depends on whether ln or ln -s is used. To safely create links using $(LN_S), either find out which form is used and adjust the arguments, or always invoke ln in the directory where the link is to be created.

In other words, it does not work to do:

          $(LN_S) foo /x/bar

Instead, do:

          (cd /x && $(LN_S) foo bar)

Set output variable RANLIB to ranlib if ranlib is found, and otherwise to : (do nothing).


If bison is found, set output variable YACC to bison -y. Otherwise, if byacc is found, set YACC to byacc. Otherwise set YACC to yacc.

5.2.2 Generic Program and File Checks

These macros are used to find programs not covered by the “particular” test macros. If you need to check the behavior of a program as well as find out whether it is present, you have to write your own test for it (see Writing Tests). By default, these macros use the environment variable PATH. If you need to check for a program that might not be in the user's PATH, you can pass a modified path to use instead, like this:

     AC_PATH_PROG([INETD], [inetd], [/usr/libexec/inetd],

You are strongly encouraged to declare the variable passed to AC_CHECK_PROG etc. as precious, See Setting Output Variables, AC_ARG_VAR, for more details.

— Macro: AC_CHECK_PROG (variable, prog-to-check-for, value-if-found, [value-if-not-found], [path], [reject])

Check whether program prog-to-check-for exists in PATH. If it is found, set variable to value-if-found, otherwise to value-if-not-found, if given. Always pass over reject (an absolute file name) even if it is the first found in the search path; in that case, set variable using the absolute file name of the prog-to-check-for found that is not reject. If variable was already set, do nothing. Calls AC_SUBST for variable.

— Macro: AC_CHECK_PROGS (variable, progs-to-check-for, [value-if-not-found], [path])

Check for each program in the whitespace-separated list progs-to-check-for existing in the PATH. If one is found, set variable to the name of that program. Otherwise, continue checking the next program in the list. If none of the programs in the list are found, set variable to value-if-not-found; if value-if-not-found is not specified, the value of variable is not changed. Calls AC_SUBST for variable.

— Macro: AC_CHECK_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROG, but first looks for prog-to-check-for with a prefix of the host type as determined by AC_CANONICAL_HOST, followed by a dash (see Canonicalizing). For example, if the user runs configure --host=i386-gnu, then this call:

          AC_CHECK_TOOL(RANLIB, ranlib, :)

sets RANLIB to i386-gnu-ranlib if that program exists in PATH, or otherwise to ranlib if that program exists in PATH, or to : if neither program exists.

— Macro: AC_CHECK_TOOLS (variable, progs-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TOOL, each of the tools in the list progs-to-check-for are checked with a prefix of the host type as determined by AC_CANONICAL_HOST, followed by a dash (see Canonicalizing). If none of the tools can be found with a prefix, then the first one without a prefix is used. If a tool is found, set variable to the name of that program. If none of the tools in the list are found, set variable to value-if-not-found; if value-if-not-found is not specified, the value of variable is not changed. Calls AC_SUBST for variable.

— Macro: AC_PATH_PROG (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROG, but set variable to the entire path of prog-to-check-for if found.

— Macro: AC_PATH_PROGS (variable, progs-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_PROGS, but if any of progs-to-check-for are found, set variable to the entire path of the program found.

— Macro: AC_PATH_TOOL (variable, prog-to-check-for, [value-if-not-found], [path])

Like AC_CHECK_TOOL, but set variable to the entire path of the program if it is found.

5.3 Files

You might also need to check for the existence of files. Before using these macros, ask yourself whether a run-time test might not be a better solution. Be aware that, like most Autoconf macros, they test a feature of the host machine, and therefore, they die when cross-compiling.

— Macro: AC_CHECK_FILE (file, [action-if-found], [action-if-not-found])

Check whether file file exists on the native system. If it is found, execute action-if-found, otherwise do action-if-not-found, if given.

— Macro: AC_CHECK_FILES (files, [action-if-found], [action-if-not-found])

Executes AC_CHECK_FILE once for each file listed in files. Additionally, defines HAVE_file (see Standard Symbols) for each file found.

5.4 Library Files

The following macros check for the presence of certain C, C++, or Fortran library archive files.

— Macro: AC_CHECK_LIB (library, function, [action-if-found], [action-if-not-found], [other-libraries])

Depending on the current language(see Language Choice), try to ensure that the C, C++, or Fortran function function is available by checking whether a test program can be linked with the library library to get the function. library is the base name of the library; e.g., to check for -lmp, use mp as the library argument.

action-if-found is a list of shell commands to run if the link with the library succeeds; action-if-not-found is a list of shell commands to run if the link fails. If action-if-found is not specified, the default action will prepend -llibrary to LIBS and define HAVE_LIBlibrary (in all capitals). This macro is intended to support building LIBS in a right-to-left (least-dependent to most-dependent) fashion such that library dependencies are satisfied as a natural side-effect of consecutive tests. Some linkers are very sensitive to library ordering so the order in which LIBS is generated is important to reliable detection of libraries.

If linking with library results in unresolved symbols that would be resolved by linking with additional libraries, give those libraries as the other-libraries argument, separated by spaces: e.g., -lXt -lX11. Otherwise, this macro will fail to detect that library is present, because linking the test program will always fail with unresolved symbols. The other-libraries argument should be limited to cases where it is desirable to test for one library in the presence of another that is not already in LIBS.

— Macro: AC_SEARCH_LIBS (function, search-libs, [action-if-found], [action-if-not-found], [other-libraries])

Search for a library defining function if it's not already available. This equates to calling AC_LINK_IFELSE([AC_LANG_CALL([], [function])]) first with no libraries, then for each library listed in search-libs.

Add -llibrary to LIBS for the first library found to contain function, and run action-if-found. If the function is not found, run action-if-not-found.

If linking with library results in unresolved symbols that would be resolved by linking with additional libraries, give those libraries as the other-libraries argument, separated by spaces: e.g., -lXt -lX11. Otherwise, this macro will fail to detect that function is present, because linking the test program will always fail with unresolved symbols.

5.5 Library Functions

The following macros check for particular C library functions. If there is no macro specifically defined to check for a function you need, and you don't need to check for any special properties of it, then you can use one of the general function-check macros.

5.5.1 Portability of C Functions

Most usual functions can either be missing, or be buggy, or be limited on some architectures. This section tries to make an inventory of these portability issues. By definition, this list will always require additions. Please help us keeping it as complete as possible.

Did you know that, on some older hosts, exit returns int? This is because exit predates void, and there was a long tradition of it returning int.
POSIX specifies that putenv puts the given string directly in environ, but some systems make a copy of it instead (eg. glibc 2.0, or BSD). And when a copy is made, unsetenv might not free it, causing a memory leak (eg. FreeBSD 4).

POSIX specifies that putenv("FOO") removes FOO from the environment, but on some systems (eg. FreeBSD 4) this is not the case and instead unsetenv must be used.

On MINGW, a call putenv("FOO=") removes FOO from the environment, rather than inserting it with an empty value.

signal handler
Normally signal takes a handler function with a return type of void, but some old systems required int instead. Any actual int value returned is not used, this is only a difference in the function prototype demanded.

All systems we know of in current use take void. Presumably int was to support K&R C, where of course void is not available. AC_TYPE_SIGNAL (see Particular Types) can be used to establish the correct type in all cases.

The ISO C99 standard says that if the output array isn't big enough and if no other errors occur, snprintf and vsnprintf truncate the output and return the number of bytes that ought to have been produced. Some older systems return the truncated length (e.g., GNU C Library 2.0.x or irix 6.5), some a negative value (e.g., earlier GNU C Library versions), and some the buffer length without truncation (e.g., 32-bit Solaris 7). Also, some buggy older systems ignore the length and overrun the buffer (e.g., 64-bit Solaris 7).
The ISO C standard says sprintf and vsprintf return the number of bytes written, but on some old systems (SunOS 4 for instance) they return the buffer pointer instead.
On various old systems, e.g., HP-UX 9, sscanf requires that its input string be writable (though it doesn't actually change it). This can be a problem when using gcc since it normally puts constant strings in read-only memory (see Incompatibilities of GCC (Using and Porting the GNU Compiler Collection)). Apparently in some cases even having format strings read-only can be a problem.
AIX 4.3 provides a broken version which produces the following results:
          strnlen ("foobar", 0) = 0
          strnlen ("foobar", 1) = 3
          strnlen ("foobar", 2) = 2
          strnlen ("foobar", 3) = 1
          strnlen ("foobar", 4) = 0
          strnlen ("foobar", 5) = 6
          strnlen ("foobar", 6) = 6
          strnlen ("foobar", 7) = 6
          strnlen ("foobar", 8) = 6
          strnlen ("foobar", 9) = 6

_SC_PAGESIZE is standard, but some older systems (eg. HP-UX 9) have _SC_PAGE_SIZE instead. This can be tested with #ifdef.
The POSIX spec says that unlink causes the given file to be removed only after there are no more open file handles for it. Not all OS's support this behavior though. So even on systems that provide unlink, you cannot portably assume it is OK to call it on files that are open. For example, on Windows 9x and ME, such a call would fail; on DOS it could even lead to file system corruption, as the file might end up being written to after the OS has removed it.
On MINGW, unsetenv is not available, but a variable FOO can be removed with a call putenv("FOO="), as described under putenv above.
The ISO C99 standard provides va_copy for copying va_list variables. It may be available in older environments too, though possibly as __va_copy (e.g., gcc in strict C89 mode). These can be tested with #ifdef. A fallback to memcpy (&dst, &src, sizeof(va_list)) will give maximum portability.
va_list is not necessarily just a pointer. It can be a struct (e.g., gcc on Alpha), which means NULL is not portable. Or it can be an array (e.g., gcc in some PowerPC configurations), which means as a function parameter it can be effectively call-by-reference and library routines might modify the value back in the caller (e.g., vsnprintf in the GNU C Library 2.1).
Signed >>
Normally the C >> right shift of a signed type replicates the high bit, giving a so-called “arithmetic” shift. But care should be taken since the ISO C standard doesn't require that behavior. On those few processors without a native arithmetic shift (for instance Cray vector systems) zero bits may be shifted in, the same as a shift of an unsigned type.

5.5.2 Particular Function Checks

These macros check for particular C functions—whether they exist, and in some cases how they respond when given certain arguments.


Check how to get alloca. Tries to get a builtin version by checking for alloca.h or the predefined C preprocessor macros __GNUC__ and _AIX. If this macro finds alloca.h, it defines HAVE_ALLOCA_H.

If those attempts fail, it looks for the function in the standard C library. If any of those methods succeed, it defines HAVE_ALLOCA. Otherwise, it sets the output variable ALLOCA to alloca.o and defines C_ALLOCA (so programs can periodically call alloca(0) to garbage collect). This variable is separate from LIBOBJS so multiple programs can share the value of ALLOCA without needing to create an actual library, in case only some of them use the code in LIBOBJS.

This macro does not try to get alloca from the System V R3 libPW or the System V R4 libucb because those libraries contain some incompatible functions that cause trouble. Some versions do not even contain alloca or contain a buggy version. If you still want to use their alloca, use ar to extract alloca.o from them instead of compiling alloca.c.

Source files that use alloca should start with a piece of code like the following, to declare it properly. In some versions of AIX, the declaration of alloca must precede everything else except for comments and preprocessor directives. The #pragma directive is indented so that pre-ANSI C compilers will ignore it, rather than choke on it.

          /* AIX requires this to be the first thing in the file.  */
          #ifndef __GNUC__
          # if HAVE_ALLOCA_H
          #  include <alloca.h>
          # else
          #  ifdef _AIX
           #pragma alloca
          #  else
          #   ifndef alloca /* predefined by HP cc +Olibcalls */
          char *alloca ();
          #   endif
          #  endif
          # endif

If the chown function is available and works (in particular, it should accept -1 for uid and gid), define HAVE_CHOWN.


If the closedir function does not return a meaningful value, define CLOSEDIR_VOID. Otherwise, callers ought to check its return value for an error indicator.


If the error_at_line function is not found, require an AC_LIBOBJ replacement of error.


If the fnmatch function conforms to POSIX, define HAVE_FNMATCH. Detect common implementation bugs, for example, the bugs in Solaris 2.4.

Note that for historical reasons, contrary to the other specific AC_FUNC macros, AC_FUNC_FNMATCH does not replace a broken/missing fnmatch. See AC_REPLACE_FNMATCH below.


Behave like AC_REPLACE_FNMATCH (replace) but also test whether fnmatch supports GNU extensions. Detect common implementation bugs, for example, the bugs in the GNU C Library 2.1.


This macro checks for the fork and vfork functions. If a working fork is found, define HAVE_WORKING_FORK. This macro checks whether fork is just a stub by trying to run it.

If vfork.h is found, define HAVE_VFORK_H. If a working vfork is found, define HAVE_WORKING_VFORK. Otherwise, define vfork to be fork for backward compatibility with previous versions of autoconf. This macro checks for several known errors in implementations of vfork and considers the system to not have a working vfork if it detects any of them. It is not considered to be an implementation error if a child's invocation of signal modifies the parent's signal handler, since child processes rarely change their signal handlers.

Since this macro defines vfork only for backward compatibility with previous versions of autoconf you're encouraged to define it yourself in new code:

          #if !HAVE_WORKING_VFORK
          # define vfork fork

If the fseeko function is available, define HAVE_FSEEKO. Define _LARGEFILE_SOURCE if necessary to make the prototype visible on some systems (e.g. glibc 2.2). Otherwise linkage problems may occur when compiling with AC_SYS_LARGEFILE on largefile-sensitive systems where off_t does not default to a 64bit entity.


If the getgroups function is available and works (unlike on Ultrix 4.3, where getgroups (0, 0) always fails), define HAVE_GETGROUPS. Set GETGROUPS_LIBS to any libraries needed to get that function. This macro runs AC_TYPE_GETGROUPS.


Check how to get the system load averages. To perform its tests properly, this macro needs the file getloadavg.c; therefore, be sure to set the AC_LIBOBJ replacement directory properly (see Generic Functions, AC_CONFIG_LIBOBJ_DIR).

If the system has the getloadavg function, define HAVE_GETLOADAVG, and set GETLOADAVG_LIBS to any libraries needed to get that function. Also add GETLOADAVG_LIBS to LIBS. Otherwise, require an AC_LIBOBJ replacement for getloadavg with source code in dir/getloadavg.c, and possibly define several other C preprocessor macros and output variables:

  1. Define C_GETLOADAVG.
  2. Define SVR4, DGUX, UMAX, or UMAX4_3 if on those systems.
  3. If nlist.h is found, define HAVE_NLIST_H.
  4. If struct nlist has an n_un.n_name member, define HAVE_STRUCT_NLIST_N_UN_N_NAME. The obsolete symbol NLIST_NAME_UNION is still defined, but do not depend upon it.
  5. Programs may need to be installed setgid (or setuid) for getloadavg to work. In this case, define GETLOADAVG_PRIVILEGED, set the output variable NEED_SETGID to true (and otherwise to false), and set KMEM_GROUP to the name of the group that should own the installed program.


Check for getmntent in the sun, seq, and gen libraries, for irix 4, PTX, and Unixware, respectively. Then, if getmntent is available, define HAVE_GETMNTENT.


Define GETPGRP_VOID if it is an error to pass 0 to getpgrp; this is the POSIX behavior. On older BSD systems, you must pass 0 to getpgrp, as it takes an argument and behaves like POSIX's getpgid.

          #if GETPGRP_VOID
            pid = getpgrp ();
            pid = getpgrp (0);

This macro does not check whether getpgrp exists at all; if you need to work in that situation, first call AC_CHECK_FUNC for getpgrp.


If link is a symbolic link, then lstat should treat link/ the same as link/.. However, many older lstat implementations incorrectly ignore trailing slashes.

It is safe to assume that if lstat incorrectly ignores trailing slashes, then other symbolic-link-aware functions like unlink also incorrectly ignore trailing slashes.

If lstat behaves properly, define LSTAT_FOLLOWS_SLASHED_SYMLINK, otherwise require an AC_LIBOBJ replacement of lstat.


If the malloc function is compatible with the GNU C library malloc (i.e., malloc (0) returns a valid pointer), define HAVE_MALLOC to 1. Otherwise define HAVE_MALLOC to 0, ask for an AC_LIBOBJ replacement for malloc, and define malloc to rpl_malloc so that the native malloc is not used in the main project.

Typically, the replacement file malloc.c should look like (note the #undef malloc):

     #if HAVE_CONFIG_H
     # include <config.h>
     #undef malloc
     #include <sys/types.h>
     void *malloc ();
     /* Allocate an N-byte block of memory from the heap.
        If N is zero, allocate a 1-byte block.  */
     void *
     rpl_malloc (size_t n)
       if (n == 0)
         n = 1;
       return malloc (n);


If the memcmp function is not available, or does not work on 8-bit data (like the one on SunOS 4.1.3), or fails when comparing 16 bytes or more and with at least one buffer not starting on a 4-byte boundary (such as the one on NeXT x86 OpenStep), require an AC_LIBOBJ replacement for memcmp.


Define HAVE_MBRTOWC to 1 if the function mbrtowc and the type mbstate_t are properly declared.


If the mktime function is not available, or does not work correctly, require an AC_LIBOBJ replacement for mktime. For the purposes of this test, mktime should conform to the POSIX standard and should be the inverse of localtime.


If the mmap function exists and works correctly, define HAVE_MMAP. Only checks private fixed mapping of already-mapped memory.


If the obstacks are found, define HAVE_OBSTACK, else require an AC_LIBOBJ replacement for obstack.


If the realloc function is compatible with the GNU C library realloc (i.e., realloc (0, 0) returns a valid pointer), define HAVE_REALLOC to 1. Otherwise define HAVE_REALLOC to 0, ask for an AC_LIBOBJ replacement for realloc, and define realloc to rpl_realloc so that the native realloc is not used in the main project. See AC_FUNC_MALLOC for details.


Determines the correct type to be passed for each of the select function's arguments, and defines those types in SELECT_TYPE_ARG1, SELECT_TYPE_ARG234, and SELECT_TYPE_ARG5 respectively. SELECT_TYPE_ARG1 defaults to int, SELECT_TYPE_ARG234 defaults to int *, and SELECT_TYPE_ARG5 defaults to struct timeval *.


If setpgrp takes no argument (the POSIX version), define SETPGRP_VOID. Otherwise, it is the BSD version, which takes two process IDs as arguments. This macro does not check whether setpgrp exists at all; if you need to work in that situation, first call AC_CHECK_FUNC for setpgrp.


Determine whether stat or lstat have the bug that it succeeds when given the zero-length file name as argument. The stat and lstat from SunOS 4.1.4 and the Hurd (as of 1998-11-01) do this.

If it does, then define HAVE_STAT_EMPTY_STRING_BUG (or HAVE_LSTAT_EMPTY_STRING_BUG) and ask for an AC_LIBOBJ replacement of it.


If setvbuf takes the buffering type as its second argument and the buffer pointer as the third, instead of the other way around, define SETVBUF_REVERSED.


If the strcoll function exists and works correctly, define HAVE_STRCOLL. This does a bit more than AC_CHECK_FUNCS(strcoll), because some systems have incorrect definitions of strcoll that should not be used.


If the strtod function does not exist or doesn't work correctly, ask for an AC_LIBOBJ replacement of strtod. In this case, because strtod.c is likely to need pow, set the output variable POW_LIB to the extra library needed.


If strerror_r is available, define HAVE_STRERROR_R, and if it is declared, define HAVE_DECL_STRERROR_R. If it returns a char * message, define STRERROR_R_CHAR_P; otherwise it returns an int error number. The Thread-Safe Functions option of POSIX requires strerror_r to return int, but many systems (including, for example, version 2.2.4 of the GNU C Library) return a char * value that is not necessarily equal to the buffer argument.


Check for strftime in the intl library, for SCO unix. Then, if strftime is available, define HAVE_STRFTIME.


If the strnlen function is not available, or is buggy (like the one from AIX 4.3), require an AC_LIBOBJ replacement for it.


If utime(file, NULL) sets file's timestamp to the present, define HAVE_UTIME_NULL.


If vprintf is found, define HAVE_VPRINTF. Otherwise, if _doprnt is found, define HAVE_DOPRNT. (If vprintf is available, you may assume that vfprintf and vsprintf are also available.)


If the fnmatch function does not conform to POSIX (see AC_FUNC_FNMATCH), ask for its AC_LIBOBJ replacement.

The files fnmatch.c, fnmatch_loop.c, and fnmatch_.h in the AC_LIBOBJ replacement directory are assumed to contain a copy of the source code of GNU fnmatch. If necessary, this source code is compiled as an AC_LIBOBJ replacement, and the fnmatch_.h file is linked to fnmatch.h so that it can be included in place of the system <fnmatch.h>.

5.5.3 Generic Function Checks

These macros are used to find functions not covered by the “particular” test macros. If the functions might be in libraries other than the default C library, first call AC_CHECK_LIB for those libraries. If you need to check the behavior of a function as well as find out whether it is present, you have to write your own test for it (see Writing Tests).

— Macro: AC_CHECK_FUNC (function, [action-if-found], [action-if-not-found])

If C function function is available, run shell commands action-if-found, otherwise action-if-not-found. If you just want to define a symbol if the function is available, consider using AC_CHECK_FUNCS instead. This macro checks for functions with C linkage even when AC_LANG(C++) has been called, since C is more standardized than C++. (see Language Choice, for more information about selecting the language for checks.)

— Macro: AC_CHECK_FUNCS (function..., [action-if-found], [action-if-not-found])

For each function in the whitespace-separated argument list, define HAVE_function (in all capitals) if it is available. If action-if-found is given, it is additional shell code to execute when one of the functions is found. You can give it a value of break to break out of the loop on the first match. If action-if-not-found is given, it is executed when one of the functions is not found.

Autoconf follows a philosophy that was formed over the years by those who have struggled for portability: isolate the portability issues in specific files, and then program as if you were in a POSIX environment. Some functions may be missing or unfixable, and your package must be ready to replace them.
— Macro: AC_LIBOBJ (function)

Specify that function.c must be included in the executables to replace a missing or broken implementation of function.

Technically, it adds function.$ac_objext to the output variable LIBOBJS if it is not already in, and calls AC_LIBSOURCE for function.c. You should not directly change LIBOBJS, since this is not traceable.

— Macro: AC_LIBSOURCE (file)

Specify that file might be needed to compile the project. If you need to know what files might be needed by a, you should trace AC_LIBSOURCE. file must be a literal.

This macro is called automatically from AC_LIBOBJ, but you must call it explicitly if you pass a shell variable to AC_LIBOBJ. In that case, since shell variables cannot be traced statically, you must pass to AC_LIBSOURCE any possible files that the shell variable might cause AC_LIBOBJ to need. For example, if you want to pass a variable $foo_or_bar to AC_LIBOBJ that holds either "foo" or "bar", you should do:


There is usually a way to avoid this, however, and you are encouraged to simply call AC_LIBOBJ with literal arguments.

Note that this macro replaces the obsolete AC_LIBOBJ_DECL, with slightly different semantics: the old macro took the function name, e.g., foo, as its argument rather than the file name.

— Macro: AC_LIBSOURCES (files)

Like AC_LIBSOURCE, but accepts one or more files in a comma-separated M4 list. Thus, the above example might be rewritten:

          AC_LIBSOURCES([foo.c, bar.c])
— Macro: AC_CONFIG_LIBOBJ_DIR (directory)

Specify that AC_LIBOBJ replacement files are to be found in directory, a relative path starting from the top level of the source tree. The replacement directory defaults to ., the top level directory, and the most typical value is lib, corresponding to AC_CONFIG_LIBOBJ_DIR(lib).

configure might need to know the replacement directory for the following reasons: (i) some checks use the replacement files, (ii) some macros bypass broken system headers by installing links to the replacement headers, etc.

It is common to merely check for the existence of a function, and ask for its AC_LIBOBJ replacement if missing. The following macro is a convenient shorthand.
— Macro: AC_REPLACE_FUNCS (function...)

Like AC_CHECK_FUNCS, but uses AC_LIBOBJ(function) as action-if-not-found. You can declare your replacement function by enclosing the prototype in #if !HAVE_function. If the system has the function, it probably declares it in a header file you should be including, so you shouldn't redeclare it lest your declaration conflict.

5.6 Header Files

The following macros check for the presence of certain C header files. If there is no macro specifically defined to check for a header file you need, and you don't need to check for any special properties of it, then you can use one of the general header-file check macros.

5.6.1 Portability of Headers

This section tries to collect knowledge about common headers, and the problems they cause. By definition, this list will always require additions. Please help us keeping it as complete as possible.

inttypes.h vs. stdint.h
Paul Eggert notes that: ISO C 1999 says that inttypes.h includes stdint.h, so there's no need to include stdint.h separately in a standard environment. Many implementations have inttypes.h but not stdint.h (e.g., Solaris 7), but I don't know of any implementation that has stdint.h but not inttypes.h. Nor do I know of any free software that includes stdint.h; stdint.h seems to be a creation of the committee.
It requires linux/types.h and sys/socket.h.
It requires linux/types.h.
On Darwin, this file requires that sys/socket.h be included beforehand. One should run:
          AC_CHECK_HEADERS([net/if.h], [], [],
          [#include <stdio.h>
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #if HAVE_SYS_SOCKET_H
          # include <sys/socket.h>

On Darwin, this file requires that stdio.h and sys/socket.h be included beforehand. One should run:
          AC_CHECK_HEADERS([netinet/if_ether.h], [], [],
          [#include <stdio.h>
          #if STDC_HEADERS
          # include <stdlib.h>
          # include <stddef.h>
          # if HAVE_STDLIB_H
          #  include <stdlib.h>
          # endif
          #if HAVE_SYS_SOCKET_H
          # include <sys/socket.h>

See above, item inttypes.h vs. stdint.h.
On many systems (e.g., Darwin), stdio.h is a prerequisite.
On FreeBSD 4.8 on ia32 and using gcc version 2.95.4, sys/params.h is a prerequisite.
On Darwin, stdlib.h is a prerequisite.
On HP Tru64 5.1, sys/types.h is a prerequisite.
Using XFree86, this header requires X11/Xlib.h, which is probably so required that you might not even consider looking for it.
          AC_CHECK_HEADERS([X11/extensions/scrnsaver.h], [], [],
          [[#include <X11/Xlib.h>

5.6.2 Particular Header Checks

These macros check for particular system header files—whether they exist, and in some cases whether they declare certain symbols.


Check for the following header files. For the first one that is found and defines DIR, define the listed C preprocessor macro:

dirent.h HAVE_DIRENT_H
sys/ndir.h HAVE_SYS_NDIR_H
sys/dir.h HAVE_SYS_DIR_H
ndir.h HAVE_NDIR_H

The directory-library declarations in your source code should look something like the following:

          #if HAVE_DIRENT_H
          # include <dirent.h>
          # define NAMLEN(dirent) strlen((dirent)->d_name)
          # define dirent direct
          # define NAMLEN(dirent) (dirent)->d_namlen
          # if HAVE_SYS_NDIR_H
          #  include <sys/ndir.h>
          # endif
          # if HAVE_SYS_DIR_H
          #  include <sys/dir.h>
          # endif
          # if HAVE_NDIR_H
          #  include <ndir.h>
          # endif

Using the above declarations, the program would declare variables to be of type struct dirent, not struct direct, and would access the length of a directory entry name by passing a pointer to a struct dirent to the NAMLEN macro.

This macro also checks for the SCO Xenix dir and x libraries.


If sys/types.h does not define major, minor, and makedev, but sys/mkdev.h does, define MAJOR_IN_MKDEV; otherwise, if sys/sysmacros.h does, define MAJOR_IN_SYSMACROS.


If the macros S_ISDIR, S_ISREG, etc. defined in sys/stat.h do not work properly (returning false positives), define STAT_MACROS_BROKEN. This is the case on Tektronix UTekV, Amdahl UTS and Motorola System V/88.


If stdbool.h exists and is conformant to C99, define HAVE_STDBOOL_H to 1; if the type _Bool is defined, define HAVE__BOOL to 1. To fulfill the C99 requirements, your system.h should contain the following code:

     # include <stdbool.h>
     # if ! HAVE__BOOL
     #  ifdef __cplusplus
     typedef bool _Bool;
     #  else
     typedef unsigned char _Bool;
     #  endif
     # endif
     # define bool _Bool
     # define false 0
     # define true 1
     # define __bool_true_false_are_defined 1


Define STDC_HEADERS if the system has ANSI C header files. Specifically, this macro checks for stdlib.h, stdarg.h, string.h, and float.h; if the system has those, it probably has the rest of the ANSI C header files. This macro also checks whether string.h declares memchr (and thus presumably the other mem functions), whether stdlib.h declare free (and thus presumably malloc and other related functions), and whether the ctype.h macros work on characters with the high bit set, as ANSI C requires.

Use STDC_HEADERS instead of __STDC__ to determine whether the system has ANSI-compliant header files (and probably C library functions) because many systems that have GCC do not have ANSI C header files.

On systems without ANSI C headers, there is so much variation that it is probably easier to declare the functions you use than to figure out exactly what the system header files declare. Some systems contain a mix of functions from ANSI and BSD; some are mostly ANSI but lack memmove; some define the BSD functions as macros in string.h or strings.h; some have only the BSD functions but string.h; some declare the memory functions in memory.h, some in string.h; etc. It is probably sufficient to check for one string function and one memory function; if the library has the ANSI versions of those then it probably has most of the others. If you put the following in

          AC_CHECK_FUNCS(strchr memcpy)

then, in your code, you can use declarations like this:

          #if STDC_HEADERS
          # include <string.h>
          # if !HAVE_STRCHR
          #  define strchr index
          #  define strrchr rindex
          # endif
          char *strchr (), *strrchr ();
          # if !HAVE_MEMCPY
          #  define memcpy(d, s, n) bcopy ((s), (d), (n))
          #  define memmove(d, s, n) bcopy ((s), (d), (n))
          # endif

If you use a function like memchr, memset, strtok, or strspn, which have no BSD equivalent, then macros won't suffice; you must provide an implementation of each function. An easy way to incorporate your implementations only when needed (since the ones in system C libraries may be hand optimized) is to, taking memchr for example, put it in memchr.c and use AC_REPLACE_FUNCS(memchr).


If sys/wait.h exists and is compatible with POSIX, define HAVE_SYS_WAIT_H. Incompatibility can occur if sys/wait.h does not exist, or if it uses the old BSD union wait instead of int to store a status value. If sys/wait.h is not POSIX compatible, then instead of including it, define the POSIX macros with their usual interpretations. Here is an example:

          #include <sys/types.h>
          #if HAVE_SYS_WAIT_H
          # include <sys/wait.h>
          #ifndef WEXITSTATUS
          # define WEXITSTATUS(stat_val) ((unsigned)(stat_val) >> 8)
          #ifndef WIFEXITED
          # define WIFEXITED(stat_val) (((stat_val) & 255) == 0)

_POSIX_VERSION is defined when unistd.h is included on POSIX systems. If there is no unistd.h, it is definitely not a POSIX system. However, some non-POSIX systems do have unistd.h.

The way to check if the system supports POSIX is:

     #if HAVE_UNISTD_H
     # include <sys/types.h>
     # include <unistd.h>
     #ifdef _POSIX_VERSION
     /* Code for POSIX systems.  */

If a program may include both time.h and sys/time.h, define TIME_WITH_SYS_TIME. On some older systems, sys/time.h includes time.h, but time.h is not protected against multiple inclusion, so programs should not explicitly include both files. This macro is useful in programs that use, for example, struct timeval as well as struct tm. It is best used in conjunction with HAVE_SYS_TIME_H, which can be checked for using AC_CHECK_HEADERS(sys/time.h).

          #if TIME_WITH_SYS_TIME
          # include <sys/time.h>
          # include <time.h>
          # if HAVE_SYS_TIME_H
          #  include <sys/time.h>
          # else
          #  include <time.h>
          # endif

If the use of TIOCGWINSZ requires <sys/ioctl.h>, then define GWINSZ_IN_SYS_IOCTL. Otherwise TIOCGWINSZ can be found in <termios.h>.


          #if HAVE_TERMIOS_H
          # include <termios.h>
          #if GWINSZ_IN_SYS_IOCTL
          # include <sys/ioctl.h>

5.6.3 Generic Header Checks

These macros are used to find system header files not covered by the “particular” test macros. If you need to check the contents of a header as well as find out whether it is present, you have to write your own test for it (see Writing Tests).

— Macro: AC_CHECK_HEADER (header-file, [action-if-found], [action-if-not-found], [includes = default-includes])

If the system header file header-file is compilable, execute shell commands action-if-found, otherwise execute action-if-not-found. If you just want to define a symbol if the header file is available, consider using AC_CHECK_HEADERS instead.

For compatibility issues with older versions of Autoconf, please read below.

— Macro: AC_CHECK_HEADERS (header-file..., [action-if-found], [action-if-not-found], [includes = default-includes])

For each given system header file header-file in the whitespace-separated argument list that exists, define HAVE_header-file (in all capitals). If action-if-found is given, it is additional shell code to execute when one of the header files is found. You can give it a value of break to break out of the loop on the first match. If action-if-not-found is given, it is executed when one of the header files is not found.

For compatibility issues with older versions of Autoconf, please read below.

Previous versions of Autoconf merely checked whether the header was accepted by the preprocessor. This was changed because the old test was inappropriate for typical uses. Headers are typically used to compile, not merely to preprocess, and the old behavior sometimes accepted headers that clashed at compile-time. If you need to check whether a header is preprocessable, you can use AC_PREPROC_IFELSE (see Running the Preprocessor).

This scheme, which improves the robustness of the test, also requires that you make sure that headers that must be included before the header-file be part of the includes, (see Default Includes). If looking for bar.h, which requires that foo.h be included before if it exists, we suggest the following scheme:

AC_CHECK_HEADERS([bar.h], [], [],
# include <foo.h>
# endif

5.7 Declarations

The following macros check for the declaration of variables and functions. If there is no macro specifically defined to check for a symbol you need, then you can use the general macros (see Generic Declarations) or, for more complex tests, you may use AC_COMPILE_IFELSE (see Running the Compiler).

5.7.1 Particular Declaration Checks

There are no specific macros for declarations.

5.7.2 Generic Declaration Checks

These macros are used to find declarations not covered by the “particular” test macros.

— Macro: AC_CHECK_DECL (symbol, [action-if-found], [action-if-not-found], [includes = default-includes])

If symbol (a function or a variable) is not declared in includes and a declaration is needed, run the shell commands action-if-not-found, otherwise action-if-found. If no includes are specified, the default includes are used (see Default Includes).

This macro actually tests whether it is valid to use symbol as an r-value, not if it is really declared, because it is much safer to avoid introducing extra declarations when they are not needed.

— Macro: AC_CHECK_DECLS (symbols, [action-if-found], [action-if-not-found], [includes = default-includes])

For each of the symbols (comma-separated list), define HAVE_DECL_symbol (in all capitals) to 1 if symbol is declared, otherwise to 0. If action-if-not-found is given, it is additional shell code to execute when one of the function declarations is needed, otherwise action-if-found is executed.

This macro uses an m4 list as first argument:

          AC_CHECK_DECLS([malloc, realloc, calloc, free])

Unlike the other AC_CHECK_*S macros, when a symbol is not declared, HAVE_DECL_symbol is defined to 0 instead of leaving HAVE_DECL_symbol undeclared. When you are sure that the check was performed, use HAVE_DECL_symbol just like any other result of Autoconf:

          #if !HAVE_DECL_SYMBOL
          extern char *symbol;

If the test may have not been performed, however, because it is safer not to declare a symbol than to use a declaration that conflicts with the system's one, you should use:

          #if defined HAVE_DECL_MALLOC && !HAVE_DECL_MALLOC
          void *malloc (size_t *s);

You fall into the second category only in extreme situations: either your files may be used without being configured, or they are used during the configuration. In most cases the traditional approach is enough.

5.8 Structures

The following macros check for the presence of certain members in C structures. If there is no macro specifically defined to check for a member you need, then you can use the general structure-member macros (see Generic Structures) or, for more complex tests, you may use AC_COMPILE_IFELSE (see Running the Compiler).

5.8.1 Particular Structure Checks

The following macros check for certain structures or structure members.


If struct stat contains an st_blksize member, define HAVE_STRUCT_STAT_ST_BLKSIZE. The former name, HAVE_ST_BLKSIZE is to be avoided, as its support will cease in the future. This macro is obsoleted, and should be replaced by

          AC_CHECK_MEMBERS([struct stat.st_blksize])

If struct stat contains an st_blocks member, define HAVE_STRUCT_STAT_ST_BLOCKS. Otherwise, require an AC_LIBOBJ replacement of fileblocks. The former name, HAVE_ST_BLOCKS is to be avoided, as its support will cease in the future.


If struct stat contains an st_rdev member, define HAVE_STRUCT_STAT_ST_RDEV. The former name for this macro, HAVE_ST_RDEV, is to be avoided as it will cease to be supported in the future. Actually, even the new macro is obsolete and should be replaced by:

          AC_CHECK_MEMBERS([struct stat.st_rdev])

If time.h does not define struct tm, define TM_IN_SYS_TIME, which means that including sys/time.h had better define struct tm.


Figure out how to get the current timezone. If struct tm has a tm_zone member, define HAVE_STRUCT_TM_TM_ZONE (and the obsoleted HAVE_TM_ZONE). Otherwise, if the external array tzname is found, define HAVE_TZNAME.

5.8.2 Generic Structure Checks

These macros are used to find structure members not covered by the “particular” test macros.

— Macro: AC_CHECK_MEMBER (aggregate.member, [action-if-found], [action-if-not-found], [includes = default-includes])

Check whether member is a member of the aggregate aggregate. If no includes are specified, the default includes are used (see Default Includes).

          AC_CHECK_MEMBER(struct passwd.pw_gecos,,
                          [AC_MSG_ERROR([We need `passwd.pw_gecos'!])],
                          [#include <pwd.h>])

You can use this macro for sub-members:

— Macro: AC_CHECK_MEMBERS (members, [action-if-found], [action-if-not-found], [includes = default-includes])

Check for the existence of each aggregate.member of members using the previous macro. When member belongs to aggregate, define HAVE_aggregate_member (in all capitals, with spaces and dots replaced by underscores). If action-if-found is given, it is executed for each of the found members. If action-if-not-found is given, it is executed for each of the members that could not be found.

This macro uses m4 lists:

          AC_CHECK_MEMBERS([struct stat.st_rdev, struct stat.st_blksize])

5.9 Types

The following macros check for C types, either builtin or typedefs. If there is no macro specifically defined to check for a type you need, and you don't need to check for any special properties of it, then you can use a general type-check macro.

5.9.1 Particular Type Checks

These macros check for particular C types in sys/types.h, stdlib.h and others, if they exist.


Define GETGROUPS_T to be whichever of gid_t or int is the base type of the array argument to getgroups.


Define HAVE_MBSTATE_T if <wchar.h> declares the mbstate_t type. Also, define mbstate_t to be a type if <wchar.h> does not declare it.


Equivalent to AC_CHECK_TYPE(mode_t, int).

— Macro: AC_TYPE_OFF_T

Equivalent to AC_CHECK_TYPE(off_t, long).

— Macro: AC_TYPE_PID_T

Equivalent to AC_CHECK_TYPE(pid_t, int).


If signal.h declares signal as returning a pointer to a function returning void, define RETSIGTYPE to be void; otherwise, define it to be int.

Define signal handlers as returning type RETSIGTYPE:

          hup_handler ()

Equivalent to AC_CHECK_TYPE(size_t, unsigned).

— Macro: AC_TYPE_UID_T

If uid_t is not defined, define uid_t to be int and gid_t to be int.

5.9.2 Generic Type Checks

These macros are used to check for types not covered by the “particular” test macros.

— Macro: AC_CHECK_TYPE (type, [action-if-found], [action-if-not-found], [includes = default-includes])

Check whether type is defined. It may be a compiler builtin type or defined by the includes (see Default Includes).

— Macro: AC_CHECK_TYPES (types, [action-if-found], [action-if-not-found], [includes = default-includes])

For each type of the types that is defined, define HAVE_type (in all capitals). If no includes are specified, the default includes are used (see Default Includes). If action-if-found is given, it is additional shell code to execute when one of the types is found. If action-if-not-found is given, it is executed when one of the types is not found.

This macro uses m4 lists:

          AC_CHECK_TYPES([unsigned long long, uintmax_t])

Autoconf, up to 2.13, used to provide to another version of AC_CHECK_TYPE, broken by design. In order to keep backward compatibility, a simple heuristics, quite safe but not totally, is implemented. In case of doubt, read the documentation of the former AC_CHECK_TYPE, see Obsolete Macros.

5.10 Compilers and Preprocessors

All the tests for compilers (AC_PROG_CC, AC_PROG_CXX, AC_PROG_F77) define the output variable EXEEXT based on the output of the compiler, typically to the empty string if Unix and .exe if Win32 or OS/2.

They also define the output variable OBJEXT based on the output of the compiler, after .c files have been excluded, typically to o if Unix, obj if Win32.

If the compiler being used does not produce executables, the tests fail. If the executables can't be run, and cross-compilation is not enabled, they fail too. See Manual Configuration, for more on support for cross compiling.

5.10.1 Specific Compiler Characteristics

Some compilers exhibit different behaviors.

Static/Dynamic Expressions
Autoconf relies on a trick to extract one bit of information from the C compiler: using negative array sizes. For instance the following excerpt of a C source demonstrates how to test whether ints are 4 bytes long:
          main (void)
            static int test_array [sizeof (int) == 4 ? 1 : -1];
            test_array [0] = 0
            return 0;

To our knowledge, there is a single compiler that does not support this trick: the HP C compilers (the real one, not only the “bundled”) on HP-UX 11.00:

          $ cc -c -Ae +O2 +Onolimit conftest.c
          cc: "conftest.c": error 1879: Variable-length arrays cannot \
              have static storage.

Autoconf works around this problem by casting sizeof (int) to long before comparing it.

5.10.2 Generic Compiler Characteristics

— Macro: AC_CHECK_SIZEOF (type, [unused], [includes = default-includes])

Define SIZEOF_type (see Standard Symbols) to be the size in bytes of type. If type is unknown, it gets a size of 0. If no includes are specified, the default includes are used (see Default Includes). If you provide include, be sure to include stdio.h which is required for this macro to run.

This macro now works even when cross-compiling. The unused argument was used when cross-compiling.

For example, the call

          AC_CHECK_SIZEOF(int *)

defines SIZEOF_INT_P to be 8 on DEC Alpha AXP systems.


Normally Autoconf ignores warnings generated by the compiler, linker, and preprocessor. If this macro is used, warnings will be treated as fatal errors instead for the current language. This macro is useful when the results of configuration will be used where warnings are unacceptable; for instance, if parts of a program are built with the GCC -Werror option. If the whole program will be built using -Werror it is often simpler to put -Werror in the compiler flags (CFLAGS etc.).

5.10.3 C Compiler Characteristics

The following macros provide ways to find and exercise a C Compiler. There are a few constructs that ought to be avoided, but do not deserve being checked for, since they can easily be worked around.

Don't use lines containing solitary backslashes
They tickle a bug in the HP-UX C compiler (checked on HP-UX 10.20, 11.00, and 11i). Running the compiler on the following source,
          #ifdef __STDC__
          * A comment with backslash-newlines in it. %{ %} *\
          char str[] = "\\
          " A string with backslash-newlines in it %{ %} \\
          char apostrophe = '\\


          error-->cpp: "foo.c", line 13: error 4048: Non-terminating comment at end of file.
          error-->cpp: "foo.c", line 13: error 4033: Missing #endif at end of file.

Removing the lines with solitary backslashes solves the problem.

Don't compile several files at once if output matters to you
Some compilers, such as the HP's, reports the name of the file it is compiling when they are several. For instance:
          $ cc a.c b.c

This can cause problems if you observe the output of the compiler to detect failures. Invoking cc -c a.c -o a.o; cc -c b.c -o b.o; cc a.o b.o -o c solves the issue.

Don't rely on correct #line support
On Solaris 8, c89 (Sun WorkShop 6 update 2 C 5.3 Patch 111679-08 2002/05/09)) rejects #line directives whose line numbers are greater than 32767. In addition, nothing in posix makes this invalid. That is the reason why Autoconf stopped issuing #line directives.
— Macro: AC_PROG_CC ([compiler-search-list])

Determine a C compiler to use. If CC is not already set in the environment, check for gcc and cc, then for other C compilers. Set output variable CC to the name of the compiler found.

This macro may, however, be invoked with an optional first argument which, if specified, must be a space separated list of C compilers to search for. This just gives the user an opportunity to specify an alternative search list for the C compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_CC like this:

          AC_PROG_CC(cl egcs gcc cc)

If the C compiler is not in ANSI C mode by default, try to add an option to output variable CC to make it so. This macro tries various options that select ANSI C on some system or another. It considers the compiler to be in ANSI C mode if it handles function prototypes correctly.

After calling this macro you can check whether the C compiler has been set to accept ANSI C; if not, the shell variable ac_cv_prog_cc_stdc is set to no. If you wrote your source code in ANSI C, you can make an un-ANSIfied copy of it by using the program ansi2knr, which comes with Automake. See also under AC_C_PROTOTYPES below.

If using the GNU C compiler, set shell variable GCC to yes. If output variable CFLAGS was not already set, set it to -g -O2 for the GNU C compiler (-O2 on systems where GCC does not accept -g), or -g for other compilers.

— Macro: AC_PROG_CC_C_O

If the C compiler does not accept the -c and -o options simultaneously, define NO_MINUS_C_MINUS_O. This macro actually tests both the compiler found by AC_PROG_CC, and, if different, the first cc in the path. The test fails if one fails. This macro was created for GNU Make to choose the default C compilation rule.

— Macro: AC_PROG_CPP

Set output variable CPP to a command that runs the C preprocessor. If $CC -E doesn't work, /lib/cpp is used. It is only portable to run CPP on files with a .c extension.

Some preprocessors don't indicate missing include files by the error status. For such preprocessors an internal variable is set that causes other macros to check the standard error from the preprocessor and consider the test failed if any warnings have been reported. For most preprocessors, though, warnings do not cause include-file tests to fail unless AC_PROG_CPP_WERROR is also specified.


This acts like AC_PROG_CPP, except it treats warnings from the preprocessor as errors even if the preprocessor exit status indicates success. This is useful for avoiding headers that generate mandatory warnings, such as deprecation notices.

The following macros check for C compiler or machine architecture features. To check for characteristics not listed here, use AC_COMPILE_IFELSE (see Running the Compiler) or AC_RUN_IFELSE (see Run Time).


Define HAVE_C_BACKSLASH_A to 1 if the C compiler understands \a.

— Macro: AC_C_BIGENDIAN ([action-if-true], [action-if-false], [action-if-unknown])

If words are stored with the most significant byte first (like Motorola and SPARC CPUs), execute action-if-true. If words are stored with the least significant byte first (like Intel and VAX CPUs), execute action-if-false.

This macro runs a test-case if endianness cannot be determined from the system header files. When cross-compiling, the test-case is not run but grep'ed for some magic values. action-if-unknown is executed if the latter case fails to determine the byte sex of the host system.

The default for action-if-true is to define WORDS_BIGENDIAN. The default for action-if-false is to do nothing. And finally, the default for action-if-unknown is to abort configure and tell the installer which variable he should preset to bypass this test.

— Macro: AC_C_CONST

If the C compiler does not fully support the ANSI C qualifier const, define const to be empty. Some C compilers that do not define __STDC__ do support const; some compilers that define __STDC__ do not completely support const. Programs can simply use const as if every C compiler supported it; for those that don't, the Makefile or configuration header file will define it as empty.

Occasionally installers use a C++ compiler to compile C code, typically because they lack a C compiler. This causes problems with const, because C and C++ treat const differently. For example:

          const int foo;

is valid in C but not in C++. These differences unfortunately cannot be papered over by defining const to be empty.

If autoconf detects this situation, it leaves const alone, as this generally yields better results in practice. However, using a C++ compiler to compile C code is not recommended or supported, and installers who run into trouble in this area should get a C compiler like GCC to compile their C code.


If the C compiler recognizes the restrict keyword, don't do anything. If it recognizes only a variant spelling (__restrict, __restrict__, or _Restrict), then define restrict to that. Otherwise, define restrict to be empty. Thus, programs may simply use restrict as if every C compiler supported it; for those that do not, the Makefile or configuration header defines it away.

Although support in C++ for the restrict keyword is not required, several C++ compilers do accept the keyword. This macro works for them, too.


If the C compiler does not understand the keyword volatile, define volatile to be empty. Programs can simply use volatile as if every C compiler supported it; for those that do not, the Makefile or configuration header will define it as empty.

If the correctness of your program depends on the semantics of volatile, simply defining it to be empty does, in a sense, break your code. However, given that the compiler does not support volatile, you are at its mercy anyway. At least your program will compile, when it wouldn't before.

In general, the volatile keyword is a feature of ANSI C, so you might expect that volatile is available only when __STDC__ is defined. However, Ultrix 4.3's native compiler does support volatile, but does not define __STDC__.

— Macro: AC_C_INLINE

If the C compiler supports the keyword inline, do nothing. Otherwise define inline to __inline__ or __inline if it accepts one of those, otherwise define inline to be empty.


If the C type char is unsigned, define __CHAR_UNSIGNED__, unless the C compiler predefines it.


If the C compiler supports a working long double type with more range or precision than the double type, define HAVE_LONG_DOUBLE.


If the C preprocessor supports the stringizing operator, define HAVE_STRINGIZE. The stringizing operator is # and is found in macros such as this:

          #define x(y) #y

If function prototypes are understood by the compiler (as determined by AC_PROG_CC), define PROTOTYPES and __PROTOTYPES. In the case the compiler does not handle prototypes, you should use ansi2knr, which comes with the Automake distribution, to unprotoize function definitions. For function prototypes, you should first define PARAMS:

          #ifndef PARAMS
          # if PROTOTYPES
          #  define PARAMS(protos) protos
          # else /* no PROTOTYPES */
          #  define PARAMS(protos) ()
          # endif /* no PROTOTYPES */

then use it this way:

          size_t my_strlen PARAMS ((const char *));

This macro also defines __PROTOTYPES; this is for the benefit of header files that cannot use macros that infringe on user name space.


Add -traditional to output variable CC if using the GNU C compiler and ioctl does not work properly without -traditional. That usually happens when the fixed header files have not been installed on an old system. Since recent versions of the GNU C compiler fix the header files automatically when installed, this is becoming a less prevalent problem.

5.10.4 C++ Compiler Characteristics

— Macro: AC_PROG_CXX ([compiler-search-list])

Determine a C++ compiler to use. Check if the environment variable CXX or CCC (in that order) is set; if so, then set output variable CXX to its value.

Otherwise, if the macro is invoked without an argument, then search for a C++ compiler under the likely names (first g++ and c++ then other names). If none of those checks succeed, then as a last resort set CXX to g++.

This macro may, however, be invoked with an optional first argument which, if specified, must be a space separated list of C++ compilers to search for. This just gives the user an opportunity to specify an alternative search list for the C++ compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_CXX like this:

          AC_PROG_CXX(cl KCC CC cxx cc++ xlC aCC c++ g++ egcs gcc)

If using the GNU C++ compiler, set shell variable GXX to yes. If output variable CXXFLAGS was not already set, set it to -g -O2 for the GNU C++ compiler (-O2 on systems where G++ does not accept -g), or -g for other compilers.


Set output variable CXXCPP to a command that runs the C++ preprocessor. If $CXX -E doesn't work, /lib/cpp is used. It is only portable to run CXXCPP on files with a .c, .C, or .cc extension.

Some preprocessors don't indicate missing include files by the error status. For such preprocessors an internal variable is set that causes other macros to check the standard error from the preprocessor and consider the test failed if any warnings have been reported. However, it is not known whether such broken preprocessors exist for C++.

5.10.5 Fortran Compiler Characteristics

The Autoconf Fortran support is divided into two categories: legacy Fortran 77 macros (F77), and modern Fortran macros (FC). The former are intended for traditional Fortran 77 code, and have output variables like F77, FFLAGS, and FLIBS. The latter are for newer programs that can (or must) compile under the newer Fortran standards, and have output variables like FC, FCFLAGS, and FCLIBS.

Except for two new macros AC_FC_SRCEXT and AC_FC_FREEFORM (see below), the FC and F77 macros behave almost identically, and so they are documented together in this section.

— Macro: AC_PROG_F77 ([compiler-search-list])

Determine a Fortran 77 compiler to use. If F77 is not already set in the environment, then check for g77 and f77, and then some other names. Set the output variable F77 to the name of the compiler found.

This macro may, however, be invoked with an optional first argument which, if specified, must be a space separated list of Fortran 77 compilers to search for. This just gives the user an opportunity to specify an alternative search list for the Fortran 77 compiler. For example, if you didn't like the default order, then you could invoke AC_PROG_F77 like this:

          AC_PROG_F77(fl32 f77 fort77 xlf g77 f90 xlf90)

If using g77 (the GNU Fortran 77 compiler), then AC_PROG_F77 will set the shell variable G77 to yes. If the output variable FFLAGS was not already set in the environment, then set it to -g -02 for g77 (or -O2 where g77 does not accept -g). Otherwise, set FFLAGS to -g for all other Fortran 77 compilers.

— Macro: AC_PROG_FC ([compiler-search-list], [dialect])

Determine a Fortran compiler to use. If FC is not already set in the environment, then dialect is a hint to indicate what Fortran dialect to search for; the default is to search for the newest available dialect. Set the output variable FC to the name of the compiler found.

By default, newer dialects are preferred over older dialects, but if dialect is specified then older dialects are preferred starting with the specified dialect. dialect can currently be one of Fortran 77, Fortran 90, or Fortran 95. However, this is only a hint of which compiler name to prefer (e.g. f90 or f95), and no attempt is made to guarantee that a particular language standard is actually supported. Thus, it is preferable that you avoid the dialect option, and use AC_PROG_FC only for code compatible with the latest Fortran standard.

This macro may, alternatively, be invoked with an optional first argument which, if specified, must be a space separated list of Fortran compilers to search for, just as in AC_PROG_F77.

If the output variable FCFLAGS was not already set in the environment, then set it to -g -02 for GNU g77 (or -O2 where g77 does not accept -g). Otherwise, set FCFLAGS to -g for all other Fortran compilers.

— Macro: AC_PROG_F77_C_O
— Macro: AC_PROG_FC_C_O

Test if the Fortran compiler accepts the options -c and -o simultaneously, and define F77_NO_MINUS_C_MINUS_O or FC_NO_MINUS_C_MINUS_O, respectively, if it does not.

The following macros check for Fortran compiler characteristics. To check for characteristics not listed here, use AC_COMPILE_IFELSE (see Running the Compiler) or AC_RUN_IFELSE (see Run Time), making sure to first set the current language to Fortran 77 or Fortran via AC_LANG(Fortran 77) or AC_LANG(Fortran) (see Language Choice).


Determine the linker flags (e.g., -L and -l) for the Fortran intrinsic and run-time libraries that are required to successfully link a Fortran program or shared library. The output variable FLIBS or FCLIBS is set to these flags (which should be include after LIBS when linking).

This macro is intended to be used in those situations when it is necessary to mix, e.g., C++ and Fortran source code in a single program or shared library (see Mixing Fortran 77 With C and C++ (GNU Automake)).

For example, if object files from a C++ and Fortran compiler must be linked together, then the C++ compiler/linker must be used for linking (since special C++-ish things need to happen at link time like calling global constructors, instantiating templates, enabling exception support, etc.).

However, the Fortran intrinsic and run-time libraries must be linked in as well, but the C++ compiler/linker doesn't know by default how to add these Fortran 77 libraries. Hence, this macro was created to determine these Fortran libraries.

The macros AC_F77_DUMMY_MAIN/AC_FC_DUMMY_MAIN or AC_F77_MAIN/AC_FC_MAIN will probably also be necessary to link C/C++ with Fortran; see below.

— Macro: AC_F77_DUMMY_MAIN ([action-if-found], [action-if-not-found])
— Macro: AC_FC_DUMMY_MAIN ([action-if-found], [action-if-not-found])

With many compilers, the Fortran libraries detected by AC_F77_LIBRARY_LDFLAGS or AC_FC_LIBRARY_LDFLAGS provide their own main entry function that initializes things like Fortran I/O, and which then calls a user-provided entry function named (say) MAIN__ to run the user's program. The AC_F77_DUMMY_MAIN/AC_FC_DUMMY_MAIN or AC_F77_MAIN/AC_FC_MAIN macro figures out how to deal with this interaction.

When using Fortran for purely numerical functions (no I/O, etc.) often one prefers to provide one's own main and skip the Fortran library initializations. In this case, however, one may still need to provide a dummy MAIN__ routine in order to prevent linking errors on some systems. AC_F77_DUMMY_MAIN or AC_FC_DUMMY_MAIN detects whether any such routine is required for linking, and what its name is; the shell variable F77_DUMMY_MAIN or FC_DUMMY_MAIN holds this name, unknown when no solution was found, and none when no such dummy main is needed.

By default, action-if-found defines F77_DUMMY_MAIN or FC_DUMMY_MAIN to the name of this routine (e.g., MAIN__) if it is required. [action-if-not-found] defaults to exiting with an error.

In order to link with Fortran routines, the user's C/C++ program should then include the following code to define the dummy main if it is needed:

          #ifdef F77_DUMMY_MAIN
          #  ifdef __cplusplus
               extern "C"
          #  endif
             int F77_DUMMY_MAIN() { return 1; }

(Replace F77 with FC for Fortran instead of Fortran 77.)

Note that this macro is called automatically from AC_F77_WRAPPERS or AC_FC_WRAPPERS; there is generally no need to call it explicitly unless one wants to change the default actions.

— Macro: AC_F77_MAIN
— Macro: AC_FC_MAIN

As discussed above, many Fortran libraries allow you to provide an entry point called (say) MAIN__ instead of the usual main, which is then called by a main function in the Fortran libraries that initializes things like Fortran I/O. The AC_F77_MAIN/AC_FC_MAIN macro detects whether it is possible to utilize such an alternate main function, and defines F77_MAIN/FC_MAIN to the name of the function. (If no alternate main function name is found, F77_MAIN/FC_MAIN is simply defined to main.)

Thus, when calling Fortran routines from C that perform things like I/O, one should use this macro and name the "main" function F77_MAIN/FC_MAIN instead of main.

— Macro: AC_F77_WRAPPERS

Defines C macros F77_FUNC(name,NAME)/FC_FUNC(name,NAME) and F77_FUNC_(name,NAME)/FC_FUNC_(name,NAME) to properly mangle the names of C/C++ identifiers, and identifiers with underscores, respectively, so that they match the name-mangling scheme used by the Fortran compiler.

Fortran is case-insensitive, and in order to achieve this the Fortran compiler converts all identifiers into a canonical case and format. To call a Fortran subroutine from C or to write a C function that is callable from Fortran, the C program must explicitly use identifiers in the format expected by the Fortran compiler. In order to do this, one simply wraps all C identifiers in one of the macros provided by AC_F77_WRAPPERS or AC_FC_WRAPPERS. For example, suppose you have the following Fortran 77 subroutine:

                subroutine foobar(x,y)
                double precision x, y
                y = 3.14159 * x

You would then declare its prototype in C or C++ as:

          #define FOOBAR_F77 F77_FUNC(foobar,FOOBAR)
          #ifdef __cplusplus
          extern "C"  /* prevent C++ name mangling */
          void FOOBAR_F77(double *x, double *y);

Note that we pass both the lowercase and uppercase versions of the function name to F77_FUNC so that it can select the right one. Note also that all parameters to Fortran 77 routines are passed as pointers (see Mixing Fortran 77 With C and C++ (GNU Automake)).

(Replace F77 with FC for Fortran instead of Fortran 77.)

Although Autoconf tries to be intelligent about detecting the name-mangling scheme of the Fortran compiler, there may be Fortran compilers that it doesn't support yet. In this case, the above code will generate a compile-time error, but some other behavior (e.g., disabling Fortran-related features) can be induced by checking whether the F77_FUNC/FC_FUNC macro is defined.

Now, to call that routine from a C program, we would do something like:

              double x = 2.7183, y;
              FOOBAR_F77(&x, &y);

If the Fortran identifier contains an underscore (e.g., foo_bar), you should use F77_FUNC_/FC_FUNC_ instead of F77_FUNC/FC_FUNC (with the same arguments). This is because some Fortran compilers mangle names differently if they contain an underscore.

— Macro: AC_F77_FUNC (name, [shellvar])
— Macro: AC_FC_FUNC (name, [shellvar])

Given an identifier name, set the shell variable shellvar to hold the mangled version name according to the rules of the Fortran linker (see also AC_F77_WRAPPERS or AC_FC_WRAPPERS). shellvar is optional; if it is not supplied, the shell variable will be simply name. The purpose of this macro is to give the caller a way to access the name-mangling information other than through the C preprocessor as above, for example, to call Fortran routines from some language other than C/C++.

— Macro: AC_FC_SRCEXT (ext, [action-if-success], [action-if-failure])

By default, the FC macros perform their tests using a .f extension for source-code files. Some compilers, however, only enable newer language features for appropriately named files, e.g. Fortran 90 features only for .f90 files. On the other hand, some other compilers expect all source files to end in .f and require special flags to support other filename extensions. The AC_FC_SRCEXT macro deals with both of these issues.

The AC_FC_SRCEXT tries to get the FC compiler to accept files ending with the extension .ext (i.e. ext does not contain the dot). If any special compiler flags are needed for this, it stores them in the output variable FCFLAGS_ext. This extension and these flags are then used for all subsequent FC tests (until AC_FC_SRCEXT is called again).

For example, you would use AC_FC_SRCEXT(f90) to employ the .f90 extension in future tests, and it would set a FCFLAGS_f90 output variable with any extra flags that are needed to compile such files.

The FCFLAGS_ext can not be simply absorbed into FCFLAGS, for two reasons based on the limitations of some compilers. First, only one FCFLAGS_ext can be used at a time, so files with different extensions must be compiled separately. Second, FCFLAGS_ext must appear immediately before the source-code filename when compiling. So, continuing the example above, you might compile a foo.f90 file in your Makefile with the command:

          foo.o: foo.f90
               $(FC) -c $(FCFLAGS) $(FCFLAGS_f90) foo.f90

If AC_FC_SRCEXT succeeds in compiling files with the ext extension, it calls [action-if-success] (defaults to nothing). If it fails, and cannot find a way to make the FC compiler accept such files, it calls [action-if-failure] (defaults to exiting with an error message).

— Macro: AC_FC_FREEFORM ([action-if-success], [action-if-failure])

The AC_FC_FREEFORM tries to ensure that the Fortran compiler ($FC) allows free-format source code (as opposed to the older fixed-format style from Fortran 77). If necessary, it may add some additional flags to FCFLAGS.

This macro is most important if you are using the default .f extension, since many compilers interpret this extension as indicating fixed-format source unless an additional flag is supplied. If you specify a different extension with AC_FC_SRCEXT, such as .f90 or .f95, then AC_FC_FREEFORM will ordinarily succeed without modifying FCFLAGS.

If AC_FC_FREEFORM succeeds in compiling free-form source, it calls [action-if-success] (defaults to nothing). If it fails, it calls [action-if-failure] (defaults to exiting with an error message).

5.11 System Services

The following macros check for operating system services or capabilities.

— Macro: AC_PATH_X ([library = Xt], [header-file = X11/Intrinsic.h], [function-call = XtMalloc (0)])

Try to locate the X Window System include files and libraries. If the user gave the command line options --x-includes=dir and --x-libraries=dir, use those directories. If either or both were not given, get the missing values by running xmkmf on a trivial Imakefile and examining the Makefile that it produces. If that fails (such as if xmkmf is not present), look for the files in several directories where they often reside. If either method is successful, set the shell variables x_includes and x_libraries to their locations, unless they are in directories the compiler searches by default.

This macro checks that function-call compiles when header is included and that the resulting executable links against library. The ability to specify these explicitly is an experimental, Debian-specific feature. The defaults reflect what other versions of Autoconf always check.

If both methods fail, or the user gave the command line option --without-x, set the shell variable no_x to yes; otherwise set it to the empty string.

— Macro: AC_PATH_XTRA ([library = Xt], [header-file = X11/Intrinsic.h], [function-call = XtMalloc (0)])

An enhanced version of AC_PATH_X. It adds the C compiler flags that X needs to output variable X_CFLAGS, and the X linker flags to X_LIBS. Define X_DISPLAY_MISSING if X is not available.

This macro checks that function-call compiles when header is included and that the resulting executable links against library. The ability to specify these explicitly is an experimental, Debian-specific feature. The defaults reflect what other versions of Autoconf always check.

This macro also checks for special libraries that some systems need in order to compile X programs. It adds any that the system needs to output variable X_EXTRA_LIBS. And it checks for special X11R6 libraries that need to be linked with before -lX11, and adds any found to the output variable X_PRE_LIBS.


Check whether the system supports starting scripts with a line of the form #! /bin/csh to select the interpreter to use for the script. After running this macro, shell code in can check the shell variable interpval; it will be set to yes if the system supports #!, no if not.


Arrange for large-file support. On some hosts, one must use special compiler options to build programs that can access large files. Append any such options to the output variable CC. Define _FILE_OFFSET_BITS and _LARGE_FILES if necessary.

Large-file support can be disabled by configuring with the --disable-largefile option.

If you use this macro, check that your program works even when off_t is longer than long, since this is common when large-file support is enabled. For example, it is not correct to print an arbitrary off_t value X with printf ("%ld", (long) X).

The LFS introduced the fseeko and ftello functions to replace their C counterparts fseek and ftell that do not use off_t. Take care to use AC_FUNC_FSEEKO to make their prototypes available when using them and large-file support is enabled.


If the system supports file names longer than 14 characters, define HAVE_LONG_FILE_NAMES.


Check to see if the POSIX termios headers and functions are available on the system. If so, set the shell variable ac_cv_sys_posix_termios to yes. If not, set the variable to no.

5.12 UNIX Variants

The following macros check for certain operating systems that need special treatment for some programs, due to exceptional oddities in their header files or libraries. These macros are warts; they will be replaced by a more systematic approach, based on the functions they make available or the environments they provide.

— Macro: AC_AIX

If on AIX, define _ALL_SOURCE. Allows the use of some BSD functions. Should be called before any macros that run the C compiler.


If using the GNU C library, define _GNU_SOURCE. Allows the use of some GNU functions. Should be called before any macros that run the C compiler.


For interactive unix (ISC), add -lcposix to output variable LIBS if necessary for POSIX facilities. Call this after AC_PROG_CC and before any other macros that use POSIX interfaces. interactive unix is no longer sold, and Sun says that they will drop support for it on 2006-07-23, so this macro is becoming obsolescent.

— Macro: AC_MINIX

If on Minix, define _MINIX and _POSIX_SOURCE and define _POSIX_1_SOURCE to be 2. This allows the use of POSIX facilities. Should be called before any macros that run the C compiler.

6 Writing Tests

If the existing feature tests don't do something you need, you have to write new ones. These macros are the building blocks. They provide ways for other macros to check whether various kinds of features are available and report the results.

This chapter contains some suggestions and some of the reasons why the existing tests are written the way they are. You can also learn a lot about how to write Autoconf tests by looking at the existing ones. If something goes wrong in one or more of the Autoconf tests, this information can help you understand the assumptions behind them, which might help you figure out how to best solve the problem.

These macros check the output of the compiler system of the current language (see Language Choice). They do not cache the results of their tests for future use (see Caching Results), because they don't know enough about the information they are checking for to generate a cache variable name. They also do not print any messages, for the same reason. The checks for particular kinds of features call these macros and do cache their results and print messages about what they're checking for.

When you write a feature test that could be applicable to more than one software package, the best thing to do is encapsulate it in a new macro. See Writing Autoconf Macros, for how to do that.

6.1 Language Choice

Autoconf-generated configure scripts check for the C compiler and its features by default. Packages that use other programming languages (maybe more than one, e.g., C and C++) need to test features of the compilers for the respective languages. The following macros determine which programming language is used in the subsequent tests in

— Macro: AC_LANG (language)

Do compilation tests using the compiler, preprocessor, and file extensions for the specified language.

Supported languages are:

Do compilation tests using CC and CPP and use extension .c for test programs. Use compilation flags: CPPFLAGS with CPP, and both CPPFLAGS and CFLAGS with CC.
Do compilation tests using CXX and CXXCPP and use extension .C for test programs. Use compilation flags: CPPFLAGS with CXXPP, and both CPPFLAGS and CXXFLAGS with CXX.
Fortran 77
Do compilation tests using F77 and use extension .f for test programs. Use compilation flags: FFLAGS.
Do compilation tests using FC and use extension .f (or whatever has been set by AC_FC_SRCEXT) for test programs. Use compilation flags: FCFLAGS.

— Macro: AC_LANG_PUSH (language)

Remember the current language (as set by AC_LANG) on a stack, and then select the language. Use this macro and AC_LANG_POP in macros that need to temporarily switch to a particular language.

— Macro: AC_LANG_POP ([language])

Select the language that is saved on the top of the stack, as set by AC_LANG_PUSH, and remove it from the stack.

If given, language specifies the language we just quit. It is a good idea to specify it when it's known (which should be the case...), since Autoconf will detect inconsistencies.

          AC_LANG_PUSH(Fortran 77)
          # Perform some tests on Fortran 77.
          # ...
          AC_LANG_POP(Fortran 77)
— Macro: AC_LANG_ASSERT (language)

Check statically that the current language is language. You should use this in your language specific macros to avoid that they be called with an inappropriate language.

This macro runs only at autoconf time, and incurs no cost at configure time. Sadly enough and because Autoconf is a two layer language 2, the macros AC_LANG_PUSH/AC_LANG_POP cannot be “optimizing”, therefore as much as possible you ought to avoid using them to wrap your code, rather, require from the user to run the macro with a correct current language, and check it with AC_LANG_ASSERT. And anyway, that may help the user understand she is running a Fortran macro while expecting a result about her Fortran 77 compiler...


Ensure that whichever preprocessor would currently be used for tests has been found. Calls AC_REQUIRE (see Prerequisite Macros) with an argument of either AC_PROG_CPP or AC_PROG_CXXCPP, depending on which language is current.

6.2 Writing Test Programs

Autoconf tests follow is common scheme: feeding some program with some input, and most of the time, feeding a compiler with some source file. This section is dedicated to these source samples.

6.2.1 Guidelines for Test Programs

The most important rule to follow when writing testing samples is:

Look for realism.

This motto means that testing samples must be written with the same strictness as real programs are written. In particular, you should avoid “shortcuts” and simplifications.

Don't just play with the preprocessor if you want to prepare a compilation. For instance, using cpp to check if a header is functional might let your configure accept a header which will cause some compiler error. Do not hesitate checking header with other headers included before, especially required headers.

Make sure the symbols you use are properly defined, i.e., refrain for simply declaring a function yourself instead of including the proper header.

Test programs should not write anything to the standard output. They should return 0 if the test succeeds, nonzero otherwise, so that success can be distinguished easily from a core dump or other failure; segmentation violations and other failures produce a nonzero exit status. Test programs should exit, not return, from main, because on some systems (old Suns, at least) the argument to return in main is ignored.

Test programs can use #if or #ifdef to check the values of preprocessor macros defined by tests that have already run. For example, if you call AC_HEADER_STDC, then later on in you can have a test program that includes an ANSI C header file conditionally:

     # include <stdlib.h>

If a test program needs to use or create a data file, give it a name that starts with conftest, such as The configure script cleans up by running rm -rf conftest* after running test programs and if the script is interrupted.

6.2.2 Test Functions

Function declarations in test programs should have a prototype conditionalized for C++. In practice, though, test programs rarely need functions that take arguments.

     #ifdef __cplusplus
     foo (int i)
     foo (i) int i;

Functions that test programs declare should also be conditionalized for C++, which requires extern "C" prototypes. Make sure to not include any header files containing clashing prototypes.

     #ifdef __cplusplus
     extern "C" void *malloc (size_t);
     void *malloc ();

If a test program calls a function with invalid parameters (just to see whether it exists), organize the program to ensure that it never invokes that function. You can do this by calling it in another function that is never invoked. You can't do it by putting it after a call to exit, because GCC version 2 knows that exit never returns and optimizes out any code that follows it in the same block.

If you include any header files, be sure to call the functions relevant to them with the correct number of arguments, even if they are just 0, to avoid compilation errors due to prototypes. GCC version 2 has internal prototypes for several functions that it automatically inlines; for example, memcpy. To avoid errors when checking for them, either pass them the correct number of arguments or redeclare them with a different return type (such as char).

6.2.3 Generating Sources

Autoconf provides a set of macros that can be used to generate test source files. They are written to be language generic, i.e., they actually depend on the current language (see Language Choice) to “format” the output properly.

— Macro: AC_LANG_CONFTEST (source)

Save the source text in the current test source file: conftest.extension where the extension depends on the current language.

Note that the source is evaluated exactly once, like regular Autoconf macro arguments, and therefore (i) you may pass a macro invocation, (ii) if not, be sure to double quote if needed.

— Macro: AC_LANG_SOURCE (source)

Expands into the source, with the definition of all the AC_DEFINE performed so far.

For instance executing (observe the double quotation!):

     AC_INIT(Autoconf Documentation, 2.59,
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"])
        [AC_LANG_SOURCE([[const char hw[] = "Hello, World\n";]])])
     gcc -E -dD conftest.c -o -

results in:

     # 1 "conftest.c"
     # 1169 "configure"
     # 1 "confdefs.h" 1
     #define PACKAGE_NAME "Autoconf Documentation"
     #define PACKAGE_TARNAME "autoconf-documentation"
     #define PACKAGE_VERSION "2.59"
     #define PACKAGE_STRING "Autoconf Documentation 2.59"
     #define PACKAGE_BUGREPORT ""
     #define HELLO_WORLD "Hello, World\n"
     # 1170 "configure" 2
     const char hw[] = "Hello, World\n";
— Macro: AC_LANG_PROGRAM (prologue, body)

Expands into a source file which consists of the prologue, and then body as body of the main function (e.g., main in C). Since it uses AC_LANG_SOURCE, the feature of the latter are available.

For instance:

     AC_INIT(Autoconf Documentation, 2.59,
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"])
     [AC_LANG_PROGRAM([[const char hw[] = "Hello, World\n";]],
                      [[fputs (hw, stdout);]])])
     gcc -E -dD conftest.c -o -

results in:

     # 1 "conftest.c"
     # 1169 "configure"
     # 1 "confdefs.h" 1
     #define PACKAGE_NAME "Autoconf Documentation"
     #define PACKAGE_TARNAME "autoconf-documentation"
     #define PACKAGE_VERSION "2.59"
     #define PACKAGE_STRING "Autoconf Documentation 2.59"
     #define PACKAGE_BUGREPORT ""
     #define HELLO_WORLD "Hello, World\n"
     # 1170 "configure" 2
     const char hw[] = "Hello, World\n";
     main ()
     fputs (hw, stdout);
       return 0;
— Macro: AC_LANG_CALL (prologue, function)

Expands into a source file which consists of the prologue, and then a call to the function as body of the main function (e.g., main in C). Since it uses AC_LANG_PROGRAMS, the feature of the latter are available.

This function will probably be replaced in the future by a version which would enable specifying the arguments. The use of this macro is not encouraged, as it violates strongly the typing system.

— Macro: AC_LANG_FUNC_LINK_TRY (function)

Expands into a source file which consists of a pseudo use of the function as body of the main function (e.g., main in C): a simple (function pointer) assignment. Since it uses AC_LANG_PROGRAMS, the feature of the latter are available.

As AC_LANG_CALL, this macro is documented only for completeness. It is considered to be severely broken, and in the future will be removed in favor of actual function calls (with properly typed arguments).

6.3 Running the Preprocessor

Sometimes one might need to run the preprocessor on some source file. Usually it is a bad idea, as you typically need to compile your project, not merely run the preprocessor on it; therefore you certainly want to run the compiler, not the preprocessor. Resist to the temptation of following the easiest path.

Nevertheless, if you need to run the preprocessor, then use AC_PREPROC_IFELSE.

— Macro: AC_PREPROC_IFELSE (input, [action-if-true], [action-if-false])

Run the preprocessor of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

This macro uses CPPFLAGS, but not CFLAGS, because -g, -O, etc. are not valid options to many C preprocessors.

It is customary to report unexpected failures with AC_MSG_FAILURE.

For instance:

     AC_INIT(Autoconf Documentation, 2.59,
     AC_DEFINE([HELLO_WORLD], ["Hello, World\n"])
        [AC_LANG_PROGRAM([[const char hw[] = "Hello, World\n";]],
                         [[fputs (hw, stdout);]])],
        [AC_MSG_FAILURE([unexpected preprocessor failure])])

results in:

     checking for gcc... gcc
     checking for C compiler default output... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ANSI C... none needed
     checking how to run the C preprocessor... gcc -E

The macro AC_TRY_CPP (see Obsolete Macros) used to play the role of AC_PREPROC_IFELSE, but double quotes its argument, making it impossible to use it to elaborate sources. You are encouraged to get rid of your old use of the macro AC_TRY_CPP in favor of AC_PREPROC_IFELSE, but, in the first place, are you sure you need to run the preprocessor and not the compiler?
— Macro: AC_EGREP_HEADER (pattern, header-file, action-if-found, [action-if-not-found])

If the output of running the preprocessor on the system header file header-file matches the extended regular expression pattern, execute shell commands action-if-found, otherwise execute action-if-not-found.

— Macro: AC_EGREP_CPP (pattern, program, [action-if-found], [action-if-not-found])

program is the text of a C or C++ program, on which shell variable, back quote, and backslash substitutions are performed. If the output of running the preprocessor on program matches the extended regular expression pattern, execute shell commands action-if-found, otherwise execute action-if-not-found.

6.4 Running the Compiler

To check for a syntax feature of the current language's (see Language Choice) compiler, such as whether it recognizes a certain keyword, or simply to try some library feature, use AC_COMPILE_IFELSE to try to compile a small program that uses that feature.

— Macro: AC_COMPILE_IFELSE (input, [action-if-found], [action-if-not-found])

Run the compiler and compilation flags of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

It is customary to report unexpected failures with AC_MSG_FAILURE. This macro does not try to link; use AC_LINK_IFELSE if you need to do that (see Running the Linker).

6.5 Running the Linker

To check for a library, a function, or a global variable, Autoconf configure scripts try to compile and link a small program that uses it. This is unlike Metaconfig, which by default uses nm or ar on the C library to try to figure out which functions are available. Trying to link with the function is usually a more reliable approach because it avoids dealing with the variations in the options and output formats of nm and ar and in the location of the standard libraries. It also allows configuring for cross-compilation or checking a function's run-time behavior if needed. On the other hand, it can be slower than scanning the libraries once, but accuracy is more important than speed.

AC_LINK_IFELSE is used to compile test programs to test for functions and global variables. It is also used by AC_CHECK_LIB to check for libraries (see Libraries), by adding the library being checked for to LIBS temporarily and trying to link a small program.

— Macro: AC_LINK_IFELSE (input, [action-if-found], [action-if-not-found])

Run the compiler (and compilation flags) and the linker of the current language (see Language Choice) on the input, run the shell commands action-if-true on success, action-if-false otherwise. The input can be made by AC_LANG_PROGRAM and friends.

LDFLAGS and LIBS are used for linking, in addition to the current compilation flags.

It is customary to report unexpected failures with AC_MSG_FAILURE. This macro does not try to execute the program; use AC_RUN_IFELSE if you need to do that (see Run Time).

6.6 Checking Run Time Behavior

Sometimes you need to find out how a system performs at run time, such as whether a given function has a certain capability or bug. If you can, make such checks when your program runs instead of when it is configured. You can check for things like the machine's endianness when your program initializes itself.

If you really need to test for a run-time behavior while configuring, you can write a test program to determine the result, and compile and run it using AC_RUN_IFELSE. Avoid running test programs if possible, because this prevents people from configuring your package for cross-compiling.

— Macro: AC_RUN_IFELSE (input, [action-if-found], [action-if-not-found], [action-if-cross-compiling])

If program compiles and links successfully and returns an exit status of 0 when executed, run shell commands action-if-true. Otherwise, run shell commands action-if-false.

The input can be made by AC_LANG_PROGRAM and friends. LDFLAGS and LIBS are used for linking, in addition to the compilation flags of the current language (see Language Choice).

If the compiler being used does not produce executables that run on the system where configure is being run, then the test program is not run. If the optional shell commands action-if-cross-compiling are given, they are run instead. Otherwise, configure prints an error message and exits.

In the action-if-false section, the exit status of the program is available in the shell variable $?, but be very careful to limit yourself to positive values smaller than 127; bigger values should be saved into a file by the program. Note also that you have simply no guarantee that this exit status is issued by the program, or by the failure of its compilation. In other words, use this feature if sadist only, it was reestablished because the Autoconf maintainers grew tired of receiving “bug reports”.

It is customary to report unexpected failures with AC_MSG_FAILURE.

Try to provide a pessimistic default value to use when cross-compiling makes run-time tests impossible. You do this by passing the optional last argument to AC_RUN_IFELSE. autoconf prints a warning message when creating configure each time it encounters a call to AC_RUN_IFELSE with no action-if-cross-compiling argument given. You may ignore the warning, though users will not be able to configure your package for cross-compiling. A few of the macros distributed with Autoconf produce this warning message.

To configure for cross-compiling you can also choose a value for those parameters based on the canonical system name (see Manual Configuration). Alternatively, set up a test results cache file with the correct values for the host system (see Caching Results).

To provide a default for calls of AC_RUN_IFELSE that are embedded in other macros, including a few of the ones that come with Autoconf, you can test whether the shell variable cross_compiling is set to yes, and then use an alternate method to get the results instead of calling the macros.

6.7 Systemology

This section aims at presenting some systems and pointers to documentation. It may help you addressing particular problems reported by users.

The Rosetta Stone for Unix contains a lot of interesting crossed information on various Unices.

Darwin is also known as Mac OS X. Beware that the file system can be case-preserving, but case insensitive. This can cause nasty problems, since for instance the installation attempt for a package having an INSTALL file can result in make install report that nothing was to be done!

That's all dependent on whether the file system is a UFS (case sensitive) or HFS+ (case preserving). By default Apple wants you to install the OS on HFS+. Unfortunately, there are some pieces of software which really need to be built on UFS. We may want to rebuild Darwin to have both UFS and HFS+ available (and put the /local/build tree on the UFS).

QNX 4.25
QNX is a realtime operating system running on Intel architecture meant to be scalable from the small embedded systems to the hundred processor super-computer. It claims to be POSIX certified. More information is available on the QNX home page, including the QNX man pages.
The documentation of several versions of Tru64 is available in different formats.
Unix version 7
Documentation is available in the V7 Manual.

6.8 Multiple Cases

Some operations are accomplished in several possible ways, depending on the unix variant. Checking for them essentially requires a “case statement”. Autoconf does not directly provide one; however, it is easy to simulate by using a shell variable to keep track of whether a way to perform the operation has been found yet.

Here is an example that uses the shell variable fstype to keep track of whether the remaining cases need to be checked.

     AC_MSG_CHECKING([how to get file system type])
     # The order of these tests is important.
     AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statvfs.h>
     #include <sys/fstyp.h>]])],
                       [AC_DEFINE(FSTYPE_STATVFS) fstype=SVR4])
     if test $fstype = no; then
       AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statfs.h>
     #include <sys/fstyp.h>]])],
                       [AC_DEFINE(FSTYPE_USG_STATFS) fstype=SVR3])
     if test $fstype = no; then
       AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/statfs.h>
     #include <sys/vmount.h>]])]),
                       [AC_DEFINE(FSTYPE_AIX_STATFS) fstype=AIX])
     # (more cases omitted here)

7 Results of Tests

Once configure has determined whether a feature exists, what can it do to record that information? There are four sorts of things it can do: define a C preprocessor symbol, set a variable in the output files, save the result in a cache file for future configure runs, and print a message letting the user know the result of the test.

7.1 Defining C Preprocessor Symbols

A common action to take in response to a feature test is to define a C preprocessor symbol indicating the results of the test. That is done by calling AC_DEFINE or AC_DEFINE_UNQUOTED.

By default, AC_OUTPUT places the symbols defined by these macros into the output variable DEFS, which contains an option -Dsymbol=value for each symbol defined. Unlike in Autoconf version 1, there is no variable DEFS defined while configure is running. To check whether Autoconf macros have already defined a certain C preprocessor symbol, test the value of the appropriate cache variable, as in this example:

     if test "$ac_cv_func_vprintf" != yes; then

If AC_CONFIG_HEADERS has been called, then instead of creating DEFS, AC_OUTPUT creates a header file by substituting the correct values into #define statements in a template file. See Configuration Headers, for more information about this kind of output.

— Macro: AC_DEFINE (variable, value, [description])
— Macro: AC_DEFINE (variable)

Define the C preprocessor variable variable to value (verbatim). value should not contain literal newlines, and if you are not using AC_CONFIG_HEADERS it should not contain any # characters, as make tends to eat them. To use a shell variable (which you need to do in order to define a value containing the M4 quote characters [ or ]), use AC_DEFINE_UNQUOTED instead. description is only useful if you are using AC_CONFIG_HEADERS. In this case, description is put into the generated as the comment before the macro define. The following example defines the C preprocessor variable EQUATION to be the string constant "$a > $b":

          AC_DEFINE(EQUATION, "$a > $b")

If neither value nor description are given, then value defaults to 1 instead of to the empty string. This is for backwards compatibility with older versions of Autoconf, but this usage is obsolescent and may be withdrawn in future versions of Autoconf.

— Macro: AC_DEFINE_UNQUOTED (variable, value, [description])
— Macro: AC_DEFINE_UNQUOTED (variable)

Like AC_DEFINE, but three shell expansions are performed—once—on variable and value: variable expansion ($), command substitution (`), and backslash escaping (\). Single and double quote characters in the value have no special meaning. Use this macro instead of AC_DEFINE when variable or value is a shell variable. Examples:

          AC_DEFINE_UNQUOTED(config_machfile, "$machfile")
          AC_DEFINE_UNQUOTED(GETGROUPS_T, $ac_cv_type_getgroups)

Due to a syntactical bizarreness of the Bourne shell, do not use semicolons to separate AC_DEFINE or AC_DEFINE_UNQUOTED calls from other macro calls or shell code; that can cause syntax errors in the resulting configure script. Use either spaces or newlines. That is, do this:

     AC_CHECK_HEADER(elf.h, [AC_DEFINE(SVR4) LIBS="$LIBS -lelf"])

or this:

       LIBS="$LIBS -lelf"])

instead of this:

     AC_CHECK_HEADER(elf.h, [AC_DEFINE(SVR4); LIBS="$LIBS -lelf"])

7.2 Setting Output Variables

Another way to record the results of tests is to set output variables, which are shell variables whose values are substituted into files that configure outputs. The two macros below create new output variables. See Preset Output Variables, for a list of output variables that are always available.

— Macro: AC_SUBST (variable, [value])

Create an output variable from a shell variable. Make AC_OUTPUT substitute the variable variable into output files (typically one or more Makefiles). This means that AC_OUTPUT will replace instances of @variable@ in input files with the value that the shell variable variable has when AC_OUTPUT is called. This value of variable should not contain literal newlines.

If value is given, in addition assign it to variable.

— Macro: AC_SUBST_FILE (variable)

Another way to create an output variable from a shell variable. Make AC_OUTPUT insert (without substitutions) the contents of the file named by shell variable variable into output files. This means that AC_OUTPUT will replace instances of @variable@ in output files (such as with the contents of the file that the shell variable variable names when AC_OUTPUT is called. Set the variable to /dev/null for cases that do not have a file to insert.

This macro is useful for inserting Makefile fragments containing special dependencies or other make directives for particular host or target types into Makefiles. For example, could contain:


and then a could contain:


Running configure in varying environments can be extremely dangerous. If for instance the user runs CC=bizarre-cc ./configure, then the cache, config.h, and many other output files will depend upon bizarre-cc being the C compiler. If for some reason the user runs ./configure again, or if it is run via ./config.status --recheck, (See Automatic Remaking, and see config.status Invocation), then the configuration can be inconsistent, composed of results depending upon two different compilers.

Environment variables that affect this situation, such as CC above, are called precious variables, and can be declared as such by AC_ARG_VAR.

— Macro: AC_ARG_VAR (variable, description)

Declare variable is a precious variable, and include its description in the variable section of ./configure --help.

Being precious means that

7.3 Caching Results

To avoid checking for the same features repeatedly in various configure scripts (or in repeated runs of one script), configure can optionally save the results of many checks in a cache file (see Cache Files). If a configure script runs with caching enabled and finds a cache file, it reads the results of previous runs from the cache and avoids rerunning those checks. As a result, configure can then run much faster than if it had to perform all of the checks every time.

— Macro: AC_CACHE_VAL (cache-id, commands-to-set-it)

Ensure that the results of the check identified by cache-id are available. If the results of the check were in the cache file that was read, and configure was not given the --quiet or --silent option, print a message saying that the result was cached; otherwise, run the shell commands commands-to-set-it. If the shell commands are run to determine the value, the value will be saved in the cache file just before configure creates its output files. See Cache Variable Names, for how to choose the name of the cache-id variable.

The commands-to-set-it must have no side effects except for setting the variable cache-id, see below.

— Macro: AC_CACHE_CHECK (message, cache-id, commands-to-set-it)

A wrapper for AC_CACHE_VAL that takes care of printing the messages. This macro provides a convenient shorthand for the most common way to use these macros. It calls AC_MSG_CHECKING for message, then AC_CACHE_VAL with the cache-id and commands arguments, and AC_MSG_RESULT with cache-id.

The commands-to-set-it must have no side effects except for setting the variable cache-id, see below.

It is very common to find buggy macros using AC_CACHE_VAL or AC_CACHE_CHECK, because people are tempted to call AC_DEFINE in the commands-to-set-it. Instead, the code that follows the call to AC_CACHE_VAL should call AC_DEFINE, by examining the value of the cache variable. For instance, the following macro is broken:

     [AC_CACHE_CHECK([whether true(1) works], [ac_cv_shell_true_works],
                      true && ac_cv_shell_true_works=yes
                      if test $ac_cv_shell_true_works = yes; then
                        AC_DEFINE([TRUE_WORKS], 1
                                  [Define if `true(1)' works properly.])

This fails if the cache is enabled: the second time this macro is run, TRUE_WORKS will not be defined. The proper implementation is:

     [AC_CACHE_CHECK([whether true(1) works], [ac_cv_shell_true_works],
                      true && ac_cv_shell_true_works=yes])
      if test $ac_cv_shell_true_works = yes; then
                  [Define if `true(1)' works properly.])

Also, commands-to-set-it should not print any messages, for example with AC_MSG_CHECKING; do that before calling AC_CACHE_VAL, so the messages are printed regardless of whether the results of the check are retrieved from the cache or determined by running the shell commands.

7.3.1 Cache Variable Names

The names of cache variables should have the following format:


for example, ac_cv_header_stat_broken or ac_cv_prog_gcc_traditional. The parts of the variable name are:

An abbreviation for your package or organization; the same prefix you begin local Autoconf macros with, except lowercase by convention. For cache values used by the distributed Autoconf macros, this value is ac.
Indicates that this shell variable is a cache value. This string must be present in the variable name, including the leading underscore.
A convention for classifying cache values, to produce a rational naming system. The values used in Autoconf are listed in Macro Names.
Which member of the class of cache values this test applies to. For example, which function (alloca), program (gcc), or output variable (INSTALL).
Any particular behavior of the specific member that this test applies to. For example, broken or set. This part of the name may be omitted if it does not apply.

The values assigned to cache variables may not contain newlines. Usually, their values will be Boolean (yes or no) or the names of files or functions; so this is not an important restriction.

7.3.2 Cache Files

A cache file is a shell script that caches the results of configure tests run on one system so they can be shared between configure scripts and configure runs. It is not useful on other systems. If its contents are invalid for some reason, the user may delete or edit it.

By default, configure uses no cache file (technically, it uses --cache-file=/dev/null), to avoid problems caused by accidental use of stale cache files.

To enable caching, configure accepts --config-cache (or -C) to cache results in the file config.cache. Alternatively, --cache-file=file specifies that file be the cache file. The cache file is created if it does not exist already. When configure calls configure scripts in subdirectories, it uses the --cache-file argument so that they share the same cache. See Subdirectories, for information on configuring subdirectories with the AC_CONFIG_SUBDIRS macro.

config.status only pays attention to the cache file if it is given the --recheck option, which makes it rerun configure.

It is wrong to try to distribute cache files for particular system types. There is too much room for error in doing that, and too much administrative overhead in maintaining them. For any features that can't be guessed automatically, use the standard method of the canonical system type and linking files (see Manual Configuration).

The site initialization script can specify a site-wide cache file to use, instead of the usual per-program cache. In this case, the cache file will gradually accumulate information whenever someone runs a new configure script. (Running configure merges the new cache results with the existing cache file.) This may cause problems, however, if the system configuration (e.g., the installed libraries or compilers) changes and the stale cache file is not deleted.

7.3.3 Cache Checkpointing

If your configure script, or a macro called from, happens to abort the configure process, it may be useful to checkpoint the cache a few times at key points using AC_CACHE_SAVE. Doing so will reduce the amount of time it takes to re-run the configure script with (hopefully) the error that caused the previous abort corrected.


Loads values from existing cache file, or creates a new cache file if a cache file is not found. Called automatically from AC_INIT.


Flushes all cached values to the cache file. Called automatically from AC_OUTPUT, but it can be quite useful to call AC_CACHE_SAVE at key points in

For instance:

      ... AC_INIT, etc. ...
     # Checks for programs.
      ... more program checks ...
     # Checks for libraries.
     AC_CHECK_LIB(nsl, gethostbyname)
     AC_CHECK_LIB(socket, connect)
      ... more lib checks ...
     # Might abort...
     AM_PATH_GTK(1.0.2,, [AC_MSG_ERROR([GTK not in path])])
     AM_PATH_GTKMM(0.9.5,, [AC_MSG_ERROR([GTK not in path])])
      ... AC_OUTPUT, etc. ...

7.4 Printing Messages

configure scripts need to give users running them several kinds of information. The following macros print messages in ways appropriate for each kind. The arguments to all of them get enclosed in shell double quotes, so the shell performs variable and back-quote substitution on them.

These macros are all wrappers around the echo shell command. configure scripts should rarely need to run echo directly to print messages for the user. Using these macros makes it easy to change how and when each kind of message is printed; such changes need only be made to the macro definitions and all of the callers will change automatically.

To diagnose static issues, i.e., when autoconf is run, see Reporting Messages.

— Macro: AC_MSG_CHECKING (feature-description)

Notify the user that configure is checking for a particular feature. This macro prints a message that starts with checking and ends with ... and no newline. It must be followed by a call to AC_MSG_RESULT to print the result of the check and the newline. The feature-description should be something like whether the Fortran compiler accepts C++ comments or for c89.

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_RESULT (result-description)

Notify the user of the results of a check. result-description is almost always the value of the cache variable for the check, typically yes, no, or a file name. This macro should follow a call to AC_MSG_CHECKING, and the result-description should be the completion of the message printed by the call to AC_MSG_CHECKING.

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_NOTICE (message)

Deliver the message to the user. It is useful mainly to print a general description of the overall purpose of a group of feature checks, e.g.,

          AC_MSG_NOTICE([checking if stack overflow is detectable])

This macro prints nothing if configure is run with the --quiet or --silent option.

— Macro: AC_MSG_ERROR (error-description, [exit-status])

Notify the user of an error that prevents configure from completing. This macro prints an error message to the standard error output and exits configure with exit-status (1 by default). error-description should be something like invalid value $HOME for \$HOME.

The error-description should start with a lower-case letter, and “cannot” is preferred to “can't”.

— Macro: AC_MSG_FAILURE (error-description, [exit-status])

This AC_MSG_ERROR wrapper notifies the user of an error that prevents configure from completing and that additional details are provided in config.log. This is typically used when abnormal results are found during a compilation.

— Macro: AC_MSG_WARN (problem-description)

Notify the configure user of a possible problem. This macro prints the message to the standard error output; configure continues running afterward, so macros that call AC_MSG_WARN should provide a default (back-up) behavior for the situations they warn about. problem-description should be something like ln -s seems to make hard links.

8 Programming in M4

Autoconf is written on top of two layers: M4sugar, which provides convenient macros for pure M4 programming, and M4sh, which provides macros dedicated to shell script generation.

As of this version of Autoconf, these two layers are still experimental, and their interface might change in the future. As a matter of fact, anything that is not documented must not be used.

8.1 M4 Quotation

The most common problem with existing macros is an improper quotation. This section, which users of Autoconf can skip, but which macro writers must read, first justifies the quotation scheme that was chosen for Autoconf and then ends with a rule of thumb. Understanding the former helps one to follow the latter.

8.1.1 Active Characters

To fully understand where proper quotation is important, you first need to know what the special characters are in Autoconf: # introduces a comment inside which no macro expansion is performed, , separates arguments, [ and ] are the quotes themselves, and finally ( and ) (which M4 tries to match by pairs).

In order to understand the delicate case of macro calls, we first have to present some obvious failures. Below they are “obvious-ified”, but when you find them in real life, they are usually in disguise.

Comments, introduced by a hash and running up to the newline, are opaque tokens to the top level: active characters are turned off, and there is no macro expansion:

     # define([def], ine)
     =># define([def], ine)

Each time there can be a macro expansion, there is a quotation expansion, i.e., one level of quotes is stripped:

     int tab[10];
     =>int tab10;
     [int tab[10];]
     =>int tab[10];

Without this in mind, the reader will try hopelessly to use her macro array:

     define([array], [int tab[10];])
     =>int tab10;

How can you correctly output the intended results3?

8.1.2 One Macro Call

Let's proceed on the interaction between active characters and macros with this small macro, which just returns its first argument:

     define([car], [$1])

The two pairs of quotes above are not part of the arguments of define; rather, they are understood by the top level when it tries to find the arguments of define. Therefore, it is equivalent to write:

     define(car, $1)

But, while it is acceptable for a to avoid unnecessary quotes, it is bad practice for Autoconf macros which must both be more robust and also advocate perfect style.

At the top level, there are only two possibilities: either you quote or you don't:

     car(foo, bar, baz)
     [car(foo, bar, baz)]
     =>car(foo, bar, baz)

Let's pay attention to the special characters:

     error-->EOF in argument list

The closing parenthesis is hidden in the comment; with a hypothetical quoting, the top level understood it this way:


Proper quotation, of course, fixes the problem:


The reader will easily understand the following examples:

     car(foo, bar)
     car([foo, bar])
     =>foo, bar
     car((foo, bar))
     =>(foo, bar)
     car([(foo], [bar)])
     car([], [])
     car([[]], [[]])

With this in mind, we can explore the cases where macros invoke macros....

8.1.3 Quotation and Nested Macros

The examples below use the following macros:

     define([car], [$1])
     define([active], [ACT, IVE])
     define([array], [int tab[10]])

Each additional embedded macro call introduces other possible interesting quotations:

     =>ACT, IVE

In the first case, the top level looks for the arguments of car, and finds active. Because M4 evaluates its arguments before applying the macro, active is expanded, which results in:

     car(ACT, IVE)

In the second case, the top level gives active as first and only argument of car, which results in:

     =>ACT, IVE

i.e., the argument is evaluated after the macro that invokes it. In the third case, car receives [active], which results in:


exactly as we already saw above.

The example above, applied to a more realistic example, gives:

     car(int tab[10];)
     =>int tab10;
     car([int tab[10];])
     =>int tab10;
     car([[int tab[10];]])
     =>int tab[10];

Huh? The first case is easily understood, but why is the second wrong, and the third right? To understand that, you must know that after M4 expands a macro, the resulting text is immediately subjected to macro expansion and quote removal. This means that the quote removal occurs twice—first before the argument is passed to the car macro, and second after the car macro expands to the first argument.

As the author of the Autoconf macro car, you then consider it to be incorrect that your users have to double-quote the arguments of car, so you “fix” your macro. Let's call it qar for quoted car:

     define([qar], [[$1]])

and check that qar is properly fixed:

     qar([int tab[10];])
     =>int tab[10];

Ahhh! That's much better.

But note what you've done: now that the arguments are literal strings, if the user wants to use the results of expansions as arguments, she has to use an unquoted macro call:


where she wanted to reproduce what she used to do with car:

     =>ACT, IVE

Worse yet: she wants to use a macro that produces a set of cpp macros:

     define([my_includes], [#include <stdio.h>])
     =>#include <stdio.h>
     error-->EOF in argument list

This macro, qar, because it double quotes its arguments, forces its users to leave their macro calls unquoted, which is dangerous. Commas and other active symbols are interpreted by M4 before they are given to the macro, often not in the way the users expect. Also, because qar behaves differently from the other macros, it's an exception that should be avoided in Autoconf.

8.1.4 changequote is Evil

The temptation is often high to bypass proper quotation, in particular when it's late at night. Then, many experienced Autoconf hackers finally surrender to the dark side of the force and use the ultimate weapon: changequote.

The M4 builtin changequote belongs to a set of primitives that allow one to adjust the syntax of the language to adjust it to one's needs. For instance, by default M4 uses ` and ' as quotes, but in the context of shell programming (and actually of most programming languages), that's about the worst choice one can make: because of strings and back-quoted expressions in shell code (such as 'this' and `that`), because of literal characters in usual programming languages (as in '0'), there are many unbalanced ` and '. Proper M4 quotation then becomes a nightmare, if not impossible. In order to make M4 useful in such a context, its designers have equipped it with changequote, which makes it possible to choose another pair of quotes. M4sugar, M4sh, Autoconf, and Autotest all have chosen to use [ and ]. Not especially because they are unlikely characters, but because they are characters unlikely to be unbalanced.

There are other magic primitives, such as changecom to specify what syntactic forms are comments (it is common to see changecom(<!--, -->) when M4 is used to produce HTML pages), changeword and changesyntax to change other syntactic details (such as the character to denote the n-th argument, $ by default, the parenthesis around arguments etc.).

These primitives are really meant to make M4 more useful for specific domains: they should be considered like command line options: --quotes, --comments, --words, and --syntax. Nevertheless, they are implemented as M4 builtins, as it makes M4 libraries self contained (no need for additional options).

There lies the problem....

The problem is that it is then tempting to use them in the middle of an M4 script, as opposed to its initialization. This, if not carefully thought out, can lead to disastrous effects: you are changing the language in the middle of the execution. Changing and restoring the syntax is often not enough: if you happened to invoke macros in between, these macros will be lost, as the current syntax will probably not be the one they were implemented with.

8.1.5 Quadrigraphs

When writing an Autoconf macro you may occasionally need to generate special characters that are difficult to express with the standard Autoconf quoting rules. For example, you may need to output the regular expression [^[], which matches any character other than [. This expression contains unbalanced brackets so it cannot be put easily into an M4 macro.

You can work around this problem by using one of the following quadrigraphs:

Expands to nothing.

Quadrigraphs are replaced at a late stage of the translation process, after m4 is run, so they do not get in the way of M4 quoting. For example, the string ^@<:@, independently of its quotation, will appear as ^[ in the output.

The empty quadrigraph can be used:

The name @&t@ was suggested by Paul Eggert:

I should give some credit to the @&t@ pun. The & is my own invention, but the t came from the source code of the algol68c compiler, written by Steve Bourne (of Bourne shell fame), and which used mt to denote the empty string. In C, it would have looked like something like:
          char const mt[] = "";

but of course the source code was written in Algol 68.

I don't know where he got mt from: it could have been his own invention, and I suppose it could have been a common pun around the Cambridge University computer lab at the time.

8.1.6 Quotation Rule Of Thumb

To conclude, the quotation rule of thumb is:

One pair of quotes per pair of parentheses.

Never over-quote, never under-quote, in particular in the definition of macros. In the few places where the macros need to use brackets (usually in C program text or regular expressions), properly quote the arguments!

It is common to read Autoconf programs with snippets like:

     changequote(<<, >>)dnl
     <<#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
     changequote([, ])dnl
     [atoi (*tzname);], ac_cv_var_tzname=yes, ac_cv_var_tzname=no)

which is incredibly useless since AC_TRY_LINK is already double quoting, so you just need:

     [#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
                 [atoi (*tzname);],

The M4-fluent reader will note that these two examples are rigorously equivalent, since M4 swallows both the changequote(<<, >>) and << >> when it collects the arguments: these quotes are not part of the arguments!

Simplified, the example above is just doing this:

     changequote(<<, >>)dnl
     changequote([, ])dnl

instead of simply:


With macros that do not double quote their arguments (which is the rule), double-quote the (risky) literals:

     [[#include <time.h>
     #ifndef tzname /* For SGI.  */
     extern char *tzname[]; /* RS6000 and others reject char **tzname.  */
                                     [atoi (*tzname);])],

See Quadrigraphs, for what to do if you run into a hopeless case where quoting does not suffice.

When you create a configure script using newly written macros, examine it carefully to check whether you need to add more quotes in your macros. If one or more words have disappeared in the M4 output, you need more quotes. When in doubt, quote.

However, it's also possible to put on too many layers of quotes. If this happens, the resulting configure script will contain unexpanded macros. The autoconf program checks for this problem by doing grep AC_ configure.

8.2 Using autom4te

The Autoconf suite, including M4sugar, M4sh, and Autotest, in addition to Autoconf per se, heavily rely on M4. All these different uses revealed common needs factored into a layer over m4: autom4te4.

autom4te is a preprocessor that is like m4. It supports M4 extensions designed for use in tools like Autoconf.

8.2.1 Invoking autom4te

The command line arguments are modeled after M4's:

     autom4te options files

where the files are directly passed to m4. In addition to the regular expansion, it handles the replacement of the quadrigraphs (see Quadrigraphs), and of __oline__, the current line in the output. It supports an extended syntax for the files:

This file is an M4 frozen file. Note that all the previous files are ignored. See the option --melt for the rationale.
If found in the library path, the file is included for expansion, otherwise it is ignored instead of triggering a failure.

Of course, it supports the Autoconf common subset of options:
Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Report processing steps.
Don't remove the temporary files and be even more verbose.
-I dir
Also look for input files in dir. Multiple invocations accumulate.
-o file
Save output (script or trace) to file. The file - stands for the standard output.

As an extension of m4, it includes the following options:
-W category
Report the warnings related to category (which can actually be a comma separated list). See Reporting Messages, macro AC_DIAGNOSE, for a comprehensive list of categories. Special values include:
report all the warnings
report none
treats warnings as errors
disable warnings falling into category

Warnings about syntax are enabled by default, and the environment variable WARNINGS, a comma separated list of categories, is honored. autom4te -W category will actually behave as if you had run:

          autom4te --warnings=syntax,$WARNINGS,category

If you want to disable autom4te's defaults and WARNINGS, but (for example) enable the warnings about obsolete constructs, you would use -W none,obsolete.

autom4te displays a back trace for errors, but not for warnings; if you want them, just pass -W error. For instance, on this

          [AC_RUN_IFELSE([AC_LANG_PROGRAM([exit (0)])])])

you get:

          $ autom4te -l autoconf -Wcross
 warning: AC_RUN_IFELSE called without default \
          to allow cross compiling
          $ autom4te -l autoconf -Wcross,error -f
 error: AC_RUN_IFELSE called without default \
          to allow cross compiling
          acgeneral.m4:3044: AC_RUN_IFELSE is expanded from...
 INNER is expanded from...
 OUTER is expanded from...
 the top level

Do not use frozen files. Any argument file.m4f will be replaced with file.m4. This helps tracing the macros which are executed only when the files are frozen, typically m4_define. For instance, running:
          autom4te --melt 1.m4 2.m4f 3.m4 4.m4f input.m4

is roughly equivalent to running:

          m4 1.m4 2.m4 3.m4 4.m4 input.m4


          autom4te 1.m4 2.m4f 3.m4 4.m4f input.m4

is equivalent to:

          m4 --reload-state=4.m4f input.m4

Produce a frozen state file. autom4te freezing is stricter than M4's: it must produce no warnings, and no output other than empty lines (a line with whitespace is not empty) and comments (starting with #). Please, note that contrary to m4, this options takes no argument:
          autom4te 1.m4 2.m4 3.m4 --freeze --output=3.m4f

corresponds to

          m4 1.m4 2.m4 3.m4 --freeze-state=3.m4f

-m octal-mode
Set the mode of the non-traces output to octal-mode; by default 0666.

As another additional feature over m4, autom4te caches its results. GNU M4 is able to produce a regular output and traces at the same time. Traces are heavily used in the GNU Build System: autoheader uses them to build, autoreconf to determine what GNU Build System components are used, automake to “parse” etc. To save the long runs of m4, traces are cached while performing regular expansion, and conversely. This cache is (actually, the caches are) stored in the directory autom4te.cache. It can safely be removed at any moment (especially if for some reason autom4te considers it is trashed).
-C directory
Specify the name of the directory where the result should be cached. Passing an empty value disables caching. Be sure to pass a relative path name, as for the time being, global caches are not supported.
Don't cache the results.
If a cache is used, consider it obsolete (but update it anyway).

Because traces are so important to the GNU Build System, autom4te provides high level tracing features as compared to M4, and helps exploiting the cache:
-t macro[:format]
Trace the invocations of macro according to the format. Multiple --trace arguments can be used to list several macros. Multiple --trace arguments for a single macro are not cumulative; instead, you should just make format as long as needed.

The format is a regular string, with newlines if desired, and several special escape codes. It defaults to $f:$l:$n:$%. It can use the following special escapes:

The character $.
The filename from which macro is called.
The line number from which macro is called.
The depth of the macro call. This is an M4 technical detail that you probably don't want to know about.
The name of the macro.
The numth argument of the call to macro.
All the arguments passed to macro, separated by the character sep or the string separator (, by default). Each argument is quoted, i.e., enclosed in a pair of square brackets.
As above, but the arguments are not quoted.
As above, but the arguments are not quoted, all new line characters in the arguments are smashed, and the default separator is :.

The escape $% produces single-line trace outputs (unless you put newlines in the separator), while $@ and $* do not.

See autoconf Invocation, for examples of trace uses.

-p macro
Cache the traces of macro, but do not enable traces. This is especially important to save CPU cycles in the future. For instance, when invoked, autoconf preselects all the macros that autoheader, automake, autoreconf etc. will trace, so that running m4 is not needed to trace them: the cache suffices. This results in a huge speed-up.

Finally, autom4te introduces the concept of Autom4te libraries. They consists in a powerful yet extremely simple feature: sets of combined command line arguments:
-l =language
Use the language Autom4te library. Current languages include:
create M4sugar output.
create M4sh executable shell scripts.
create Autotest executable test suites.
create Autoconf executable configure scripts.
create Autoconf executable configure scripts without reading aclocal.m4.

-B dir
Prepend directory dir to the search path. This is used to include the language-specific files before any third-party macros.

As an example, if Autoconf is installed in its default location, /usr/local, running autom4te -l m4sugar foo.m4 is strictly equivalent to running autom4te --prepend-include /usr/local/share/autoconf m4sugar/m4sugar.m4f --warnings syntax foo.m4. Recursive expansion applies: running autom4te -l m4sh foo.m4 is the same as autom4te --language M4sugar m4sugar/m4sh.m4f foo.m4, i.e., autom4te --prepend-include /usr/local/share/autoconf m4sugar/m4sugar.m4f m4sugar/m4sh.m4f --mode 777 foo.m4. The definition of the languages is stored in autom4te.cfg.

8.2.2 Customizing autom4te

One can customize autom4te via ~/.autom4te.cfg (i.e., as found in the user home directory), and ./.autom4te.cfg (i.e., as found in the directory from which autom4te is run). The order is first reading autom4te.cfg, then ~/.autom4te.cfg, then ./.autom4te.cfg, and finally the command line arguments.

In these text files, comments are introduced with #, and empty lines are ignored. Customization is performed on a per-language basis, wrapped in between a begin-language: "language", end-language: "language" pair.

Customizing a language stands for appending options (see autom4te Invocation) to the current definition of the language. Options, and more generally arguments, are introduced by args: arguments. You may use the traditional shell syntax to quote the arguments.

As an example, to disable Autoconf caches (autom4te.cache) globally, include the following lines in ~/.autom4te.cfg:

## ------------------ ##
## User Preferences.  ##
## ------------------ ##

begin-language: "Autoconf"
args: --no-cache
end-language: "Autoconf"

8.3 Programming in M4sugar

M4 by itself provides only a small, but sufficient, set of all-purpose macros. M4sugar introduces additional generic macros. Its name was coined by Lars J. Aas: “Readability And Greater Understanding Stands 4 M4sugar”.

8.3.1 Redefined M4 Macros

With a few exceptions, all the M4 native macros are moved in the m4_ pseudo-namespace, e.g., M4sugar renames define as m4_define etc.

Some M4 macros are redefined, and are slightly incompatible with their native equivalent.

— Macro: dnl

This macro kept its original name: no m4_dnl is defined.

— Macro: m4_defn (macro)

Contrary to the M4 builtin, this macro fails if macro is not defined. See m4_undefine.

— Macro: m4_exit (exit-status)

This macro corresponds to m4exit.

— Macro: m4_if (comment)
— Macro: m4_if (string-1, string-2, equal, [not-equal])
— Macro: m4_if (string-1, string-2, equal, ...)

This macro corresponds to ifelse.

— Macro: m4_undefine (macro)

Contrary to the M4 builtin, this macro fails if macro is not defined. Use

          m4_ifdef([macro], [m4_undefine([macro])])

to recover the behavior of the builtin.

— Macro: m4_bpatsubst (string, regexp, [replacement])

This macro corresponds to patsubst. The name m4_patsubst is kept for future versions of M4sh, on top of GNU M4 which will provide extended regular expression syntax via epatsubst.

— Macro: m4_popdef (macro)

Contrary to the M4 builtin, this macro fails if macro is not defined. See m4_undefine.

— Macro: m4_bregexp (string, regexp, [replacement])

This macro corresponds to regexp. The name m4_regexp is kept for future versions of M4sh, on top of GNU M4 which will provide extended regular expression syntax via eregexp.

— Macro: m4_wrap (text)

This macro corresponds to m4wrap.

You are encouraged to end text with [], so that there are no risks that two consecutive invocations of m4_wrap result in an unexpected pasting of tokens, as in

          m4_define([foo], [Foo])
          m4_define([bar], [Bar])
          m4_define([foobar], [FOOBAR])

8.3.2 Evaluation Macros

The following macros give some control over the order of the evaluation by adding or removing levels of quotes. They are meant for hard-core M4 programmers.

— Macro: m4_dquote (arg1, ...)

Return the arguments as a quoted list of quoted arguments.

— Macro: m4_quote (arg1, ...)

Return the arguments as a single entity, i.e., wrap them into a pair of quotes.

The following example aims at emphasizing the difference between (i), not using these macros, (ii), using m4_quote, and (iii), using m4_dquote.

     $ cat example.m4
     # Overquote, so that quotes are visible.
     m4_define([show], [$[]1 = [$1], $[]@ = [$@]])
     show(a, b)
     show(m4_quote(a, b))
     show(m4_dquote(a, b))
     $ autom4te -l m4sugar example.m4
     $1 = a, $@ = [a],[b]
     $1 = a,b, $@ = [a,b]
     $1 = [a],[b], $@ = [[a],[b]]

8.3.3 Forbidden Patterns

M4sugar provides a means to define suspicious patterns, patterns describing tokens which should not be found in the output. For instance, if an Autoconf configure script includes tokens such as AC_DEFINE, or dnl, then most probably something went wrong (typically a macro was not evaluated because of overquotation).

M4sugar forbids all the tokens matching ^m4_ and ^dnl$.

— Macro: m4_pattern_forbid (pattern)

Declare that no token matching pattern must be found in the output. Comments are not checked; this can be a problem if, for instance, you have some macro left unexpanded after an #include. No consensus is currently found in the Autoconf community, as some people consider it should be valid to name macros in comments (which doesn't makes sense to the author of this documentation, as #-comments should document the output, not the input, documented by dnl comments).

Of course, you might encounter exceptions to these generic rules, for instance you might have to refer to $m4_flags.

— Macro: m4_pattern_allow (pattern)

Any token matching pattern is allowed, including if it matches an m4_pattern_forbid pattern.

8.4 Programming in M4sh

M4sh, pronounced “mash”, is aiming at producing portable Bourne shell scripts. This name was coined by Lars J. Aas, who notes that, according to the Webster's Revised Unabridged Dictionary (1913):

Mash \Mash\, n. [Akin to G. meisch, maisch, meische, maische, mash, wash, and prob. to AS. miscian to mix. See “Mix”.]
  1. A mass of mixed ingredients reduced to a soft pulpy state by beating or pressure....
  2. A mixture of meal or bran and water fed to animals.
  3. A mess; trouble. [Obs.] –Beau. & Fl.

For the time being, it is not mature enough to be widely used.

M4sh provides portable alternatives for some common shell constructs that unfortunately are not portable in practice.

— Macro: AS_DIRNAME (pathname)

Return the directory portion of pathname, using the algorithm required by POSIX. See Limitations of Usual Tools, for more details about what this returns and why it is more portable than the dirname command.

— Macro: AS_IF (test, [RUN-IF-TRUE], [RUN-IF-FALSE])

Run shell code TEST. If TEST exits with a zero status then run shell code RUN-IF-TRUE, else run shell code RUN-IF-FALSE, with simplifications if either RUN-IF-TRUE or RUN-IF-FALSE is empty.

— Macro: AS_MKDIR_P (filename)

Make the directory filename, including intervening directories as necessary. This is equivalent to mkdir -p filename, except that it is portable to older versions of mkdir that lack support for the -p option.

— Macro: AS_SET_CATFILE (var, dir, file)

Set the shell variable var to dir/file, but optimizing the common cases (dir or file is ., file is absolute etc.).

9 Writing Autoconf Macros

When you write a feature test that could be applicable to more than one software package, the best thing to do is encapsulate it in a new macro. Here are some instructions and guidelines for writing Autoconf macros.

9.1 Macro Definitions

Autoconf macros are defined using the AC_DEFUN macro, which is similar to the M4 builtin m4_define macro. In addition to defining a macro, AC_DEFUN adds to it some code that is used to constrain the order in which macros are called (see Prerequisite Macros).

An Autoconf macro definition looks like this:

     AC_DEFUN(macro-name, macro-body)

You can refer to any arguments passed to the macro as $1, $2, etc. See How to define new macros (GNU m4), for more complete information on writing M4 macros.

Be sure to properly quote both the macro-body and the macro-name to avoid any problems if the macro happens to have been previously defined.

Each macro should have a header comment that gives its prototype, and a brief description. When arguments have default values, display them in the prototype. For example:

     # --------------------------------------
     [{ _AC_ECHO([configure: error: $1], 2); exit m4_default([$2], 1); }])

Comments about the macro should be left in the header comment. Most other comments will make their way into configure, so just keep using # to introduce comments.

If you have some very special comments about pure M4 code, comments that make no sense in configure and in the header comment, then use the builtin dnl: it causes M4 to discard the text through the next newline.

Keep in mind that dnl is rarely needed to introduce comments; dnl is more useful to get rid of the newlines following macros that produce no output, such as AC_REQUIRE.

9.2 Macro Names

All of the Autoconf macros have all-uppercase names starting with AC_ to prevent them from accidentally conflicting with other text. All shell variables that they use for internal purposes have mostly-lowercase names starting with ac_. To ensure that your macros don't conflict with present or future Autoconf macros, you should prefix your own macro names and any shell variables they use with some other sequence. Possibilities include your initials, or an abbreviation for the name of your organization or software package.

Most of the Autoconf macros' names follow a structured naming convention that indicates the kind of feature check by the name. The macro names consist of several words, separated by underscores, going from most general to most specific. The names of their cache variables use the same convention (see Cache Variable Names, for more information on them).

The first word of the name after AC_ usually tells the category of the feature being tested. Here are the categories used in Autoconf for specific test macros, the kind of macro that you are more likely to write. They are also used for cache variables, in all-lowercase. Use them where applicable; where they're not, invent your own categories.

C language builtin features.
Declarations of C variables in header files.
Functions in libraries.
unix group owners of files.
Header files.
C libraries.
The full path names to files, including programs.
The base names of programs.
Members of aggregates.
Operating system features.
C builtin or declared types.
C variables in libraries.

After the category comes the name of the particular feature being tested. Any further words in the macro name indicate particular aspects of the feature. For example, AC_FUNC_UTIME_NULL checks the behavior of the utime function when called with a NULL pointer.

An internal macro should have a name that starts with an underscore; Autoconf internals should therefore start with _AC_. Additionally, a macro that is an internal subroutine of another macro should have a name that starts with an underscore and the name of that other macro, followed by one or more words saying what the internal macro does. For example, AC_PATH_X has internal macros _AC_PATH_X_XMKMF and _AC_PATH_X_DIRECT.

9.3 Reporting Messages

When macros statically diagnose abnormal situations, benign or fatal, they should report them using these macros. For dynamic issues, i.e., when configure is run, see Printing Messages.

— Macro: AC_DIAGNOSE (category, message)

Report message as a warning (or as an error if requested by the user) if warnings of the category are turned on. You are encouraged to use standard categories, which currently include:

messages that don't fall into one of the following categories. Use of an empty category is equivalent.
related to cross compilation issues.
use of an obsolete construct.
dubious syntactic constructs, incorrectly ordered macro calls.

— Macro: AC_WARNING (message)

Equivalent to AC_DIAGNOSE([syntax], message), but you are strongly encouraged to use a finer grained category.

— Macro: AC_FATAL (message)

Report a severe error message, and have autoconf die.

When the user runs autoconf -W error, warnings from AC_DIAGNOSE and AC_WARNING are reported as error, see autoconf Invocation.

9.4 Dependencies Between Macros

Some Autoconf macros depend on other macros having been called first in order to work correctly. Autoconf provides a way to ensure that certain macros are called if needed and a way to warn the user if macros are called in an order that might cause incorrect operation.

9.4.1 Prerequisite Macros

A macro that you write might need to use values that have previously been computed by other macros. For example, AC_DECL_YYTEXT examines the output of flex or lex, so it depends on AC_PROG_LEX having been called first to set the shell variable LEX.

Rather than forcing the user of the macros to keep track of the dependencies between them, you can use the AC_REQUIRE macro to do it automatically. AC_REQUIRE can ensure that a macro is only called if it is needed, and only called once.

— Macro: AC_REQUIRE (macro-name)

If the M4 macro macro-name has not already been called, call it (without any arguments). Make sure to quote macro-name with square brackets. macro-name must have been defined using AC_DEFUN or else contain a call to AC_PROVIDE to indicate that it has been called.

AC_REQUIRE must be used inside an AC_DEFUN'd macro; it must not be called from the top level.

AC_REQUIRE is often misunderstood. It really implements dependencies between macros in the sense that if one macro depends upon another, the latter will be expanded before the body of the former. In particular, AC_REQUIRE(FOO) is not replaced with the body of FOO. For instance, this definition of macros:

     [test "$body_temperature_in_celsius" -gt "38" &&
     [test "$hair_style" = "curly" &&
     [if date | grep '^Sat.*pm' >/dev/null 2>&1; then

with this

     if test "$dance_floor" = occupied; then
       AC_MSG_ERROR([cannot pick up here, let's move])

will not leave you with a better chance to meet a kindred soul at other times than Saturday night since it expands into:

     test "$body_temperature_in_Celsius" -gt "38" &&
     test "$hair_style" = "curly" &&
     if date | grep '^Sat.*pm' >/dev/null 2>&1; then

This behavior was chosen on purpose: (i) it prevents messages in required macros from interrupting the messages in the requiring macros; (ii) it avoids bad surprises when shell conditionals are used, as in:

     if ...; then

You are encouraged to put all AC_REQUIREs at the beginning of a macro. You can use dnl to avoid the empty lines they leave.

9.4.2 Suggested Ordering

Some macros should be run before another macro if both are called, but neither requires that the other be called. For example, a macro that changes the behavior of the C compiler should be called before any macros that run the C compiler. Many of these dependencies are noted in the documentation.

Autoconf provides the AC_BEFORE macro to warn users when macros with this kind of dependency appear out of order in a file. The warning occurs when creating configure from, not when running configure.

For example, AC_PROG_CPP checks whether the C compiler can run the C preprocessor when given the -E option. It should therefore be called after any macros that change which C compiler is being used, such as AC_PROG_CC. So AC_PROG_CC contains:

     AC_BEFORE([$0], [AC_PROG_CPP])dnl

This warns the user if a call to AC_PROG_CPP has already occurred when AC_PROG_CC is called.

— Macro: AC_BEFORE (this-macro-name, called-macro-name)

Make M4 print a warning message to the standard error output if called-macro-name has already been called. this-macro-name should be the name of the macro that is calling AC_BEFORE. The macro called-macro-name must have been defined using AC_DEFUN or else contain a call to AC_PROVIDE to indicate that it has been called.

9.5 Obsoleting Macros

Configuration and portability technology has evolved over the years. Often better ways of solving a particular problem are developed, or ad-hoc approaches are systematized. This process has occurred in many parts of Autoconf. One result is that some of the macros are now considered obsolete; they still work, but are no longer considered the best thing to do, hence they should be replaced with more modern macros. Ideally, autoupdate should replace the old macro calls with their modern implementation.

Autoconf provides a simple means to obsolete a macro.

— Macro: AU_DEFUN (old-macro, implementation, [message])

Define old-macro as implementation. The only difference with AC_DEFUN is that the user will be warned that old-macro is now obsolete.

If she then uses autoupdate, the call to old-macro will be replaced by the modern implementation. The additional message is then printed.

9.6 Coding Style

The Autoconf macros follow a strict coding style. You are encouraged to follow this style, especially if you intend to distribute your macro, either by contributing it to Autoconf itself, or via other means.

The first requirement is to pay great attention to the quotation. For more details, see Autoconf Language, and M4 Quotation.

Do not try to invent new interfaces. It is likely that there is a macro in Autoconf that resembles the macro you are defining: try to stick to this existing interface (order of arguments, default values, etc.). We are conscious that some of these interfaces are not perfect; nevertheless, when harmless, homogeneity should be preferred over creativity.

Be careful about clashes both between M4 symbols and between shell variables.

If you stick to the suggested M4 naming scheme (see Macro Names), you are unlikely to generate conflicts. Nevertheless, when you need to set a special value, avoid using a regular macro name; rather, use an “impossible” name. For instance, up to version 2.13, the macro AC_SUBST used to remember what symbols were already defined by setting AC_SUBST_symbol, which is a regular macro name. But since there is a macro named AC_SUBST_FILE, it was just impossible to AC_SUBST(FILE)! In this case, AC_SUBST(symbol) or _AC_SUBST(symbol) should have been used (yes, with the parentheses)...or better yet, high-level macros such as AC_EXPAND_ONCE.

No Autoconf macro should ever enter the user-variable name space; i.e., except for the variables that are the actual result of running the macro, all shell variables should start with ac_. In addition, small macros or any macro that is likely to be embedded in other macros should be careful not to use obvious names.

Do not use dnl to introduce comments: most of the comments you are likely to write are either header comments which are not output anyway, or comments that should make their way into configure. There are exceptional cases where you do want to comment special M4 constructs, in which case dnl is right, but keep in mind that it is unlikely.

M4 ignores the leading spaces before each argument, use this feature to indent in such a way that arguments are (more or less) aligned with the opening parenthesis of the macro being called. For instance, instead of

     AC_CACHE_CHECK(for EMX OS/2 environment,
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM(, [return __EMX__;])],
     [ac_cv_emxos2=yes], [ac_cv_emxos2=no])])


     AC_CACHE_CHECK([for EMX OS/2 environment], [ac_cv_emxos2],
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [return __EMX__;])],

or even

     AC_CACHE_CHECK([for EMX OS/2 environment],
                                                        [return __EMX__;])],

When using AC_RUN_IFELSE or any macro that cannot work when cross-compiling, provide a pessimistic value (typically no).

Feel free to use various tricks to prevent auxiliary tools, such as syntax-highlighting editors, from behaving improperly. For instance, instead of:

     m4_bpatsubst([$1], [$"])


     m4_bpatsubst([$1], [$""])

so that Emacsen do not open an endless “string” at the first quote. For the same reasons, avoid:

     test $[#] != 0

and use:

     test $[@%:@] != 0

Otherwise, the closing bracket would be hidden inside a #-comment, breaking the bracket-matching highlighting from Emacsen. Note the preferred style to escape from M4: $[1], $[@], etc. Do not escape when it is unnecessary. Common examples of useless quotation are [$]$1 (write $$1), [$]var (use $var), etc. If you add portability issues to the picture, you'll prefer ${1+"$[@]"} to "[$]@", and you'll prefer do something better than hacking Autoconf :-).

When using sed, don't use -e except for indenting purpose. With the s command, the preferred separator is / unless / itself is used in the command, in which case you should use ,.

See Macro Definitions, for details on how to define a macro. If a macro doesn't use AC_REQUIRE and it is expected to never be the object of an AC_REQUIRE directive, then use m4_define. In case of doubt, use AC_DEFUN. All the AC_REQUIRE statements should be at the beginning of the macro, dnl'ed.

You should not rely on the number of arguments: instead of checking whether an argument is missing, test that it is not empty. It provides both a simpler and a more predictable interface to the user, and saves room for further arguments.

Unless the macro is short, try to leave the closing ]) at the beginning of a line, followed by a comment that repeats the name of the macro being defined. This introduces an additional newline in configure; normally, that is not a problem, but if you want to remove it you can use []dnl on the last line. You can similarly use []dnl after a macro call to remove its newline. []dnl is recommended instead of dnl to ensure that M4 does not interpret the dnl as being attached to the preceding text or macro output. For example, instead of:

     [AC_MSG_CHECKING([for X])
     # ...omitted...
       AC_MSG_RESULT([libraries $x_libraries, headers $x_includes])

you would write:

     AC_MSG_CHECKING([for X])
     # ...omitted...
       AC_MSG_RESULT([libraries $x_libraries, headers $x_includes])
     ])# AC_PATH_X

If the macro is long, try to split it into logical chunks. Typically, macros that check for a bug in a function and prepare its AC_LIBOBJ replacement should have an auxiliary macro to perform this setup. Do not hesitate to introduce auxiliary macros to factor your code.

In order to highlight the recommended coding style, here is a macro written the old way:

     dnl Check for EMX on OS/2.
     dnl _AC_EMXOS2
     [AC_CACHE_CHECK(for EMX OS/2 environment, ac_cv_emxos2,
     ac_cv_emxos2=yes, ac_cv_emxos2=no)])
     test "$ac_cv_emxos2" = yes && EMXOS2=yes])

and the new way:

     # _AC_EMXOS2
     # ----------
     # Check for EMX on OS/2.
     [AC_CACHE_CHECK([for EMX OS/2 environment], [ac_cv_emxos2],
     [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [return __EMX__;])],
     test "$ac_cv_emxos2" = yes && EMXOS2=yes[]dnl
     ])# _AC_EMXOS2

10 Portable Shell Programming

When writing your own checks, there are some shell-script programming techniques you should avoid in order to make your code portable. The Bourne shell and upward-compatible shells like the Korn shell and Bash have evolved over the years, but to prevent trouble, do not take advantage of features that were added after unix version 7, circa 1977 (see Systemology).

You should not use shell functions, aliases, negated character classes, or other features that are not found in all Bourne-compatible shells; restrict yourself to the lowest common denominator. Even unset is not supported by all shells! Also, include a space after the exclamation point in interpreter specifications, like this:

     #! /usr/bin/perl

If you omit the space before the path, then 4.2BSD based systems (such as DYNIX) will ignore the line, because they interpret #! / as a 4-byte magic number. Some old systems have quite small limits on the length of the #! line too, for instance 32 bytes (not including the newline) on SunOS 4.

The set of external programs you should run in a configure script is fairly small. See Utilities in Makefiles (GNU Coding Standards), for the list. This restriction allows users to start out with a fairly small set of programs and build the rest, avoiding too many interdependencies between packages.

Some of these external utilities have a portable subset of features; see Limitations of Usual Tools.

There are other sources of documentation about shells. See for instance the Shell FAQs.

10.1 Shellology

There are several families of shells, most prominently the Bourne family and the C shell family which are deeply incompatible. If you want to write portable shell scripts, avoid members of the C shell family. The the Shell difference FAQ includes a small history of Unix shells, and a comparison between several of them.

Below we describe some of the members of the Bourne shell family.

ash is often used on GNU/Linux and BSD systems as a light-weight Bourne-compatible shell. Ash 0.2 has some bugs that are fixed in the 0.3.x series, but portable shell scripts should work around them, since version 0.2 is still shipped with many GNU/Linux distributions.

To be compatible with Ash 0.2:

To detect whether you are running bash, test if BASH_VERSION is set. To disable its extensions and require POSIX compatibility, run set -o posix. See Bash POSIX Mode (The GNU Bash Reference Manual), for details.
Bash 2.05 and later
Versions 2.05 and later of bash use a different format for the output of the set builtin, designed to make evaluating its output easier. However, this output is not compatible with earlier versions of bash (or with many other shells, probably). So if you use bash 2.05 or higher to execute configure, you'll need to use bash 2.05 for all other build tasks as well.
The Korn shell is compatible with the Bourne family and it mostly conforms to POSIX. It has two major variants commonly called ksh88 and ksh93, named after the years of initial release. It is usually called ksh, but Solaris systems have three variants: /usr/bin/ksh is ksh88, /usr/xpg4/bin/sh is a POSIX-compliant variant of ksh88, and /usr/dt/bin/dtksh is ksh93. /usr/bin/ksh is standard on Solaris; the other variants are parts of optional packages. There is no extra charge for these packages, but they are not part of a minimal OS install and therefore some installations may not have it. A public-domain clone of the Korn shell called pdksh is also widely available: it has most of the ksh88 features along with a few of its own.
To detect whether you are running zsh, test if ZSH_VERSION is set. By default zsh is not compatible with the Bourne shell: you have to run emulate sh and set NULLCMD to :. See Compatibility (The Z Shell Manual), for details.

Zsh 3.0.8 is the native /bin/sh on Mac OS X 10.0.3.

The following discussion between Russ Allbery and Robert Lipe is worth reading:

Russ Allbery:

The GNU assumption that /bin/sh is the one and only shell leads to a permanent deadlock. Vendors don't want to break users' existing shell scripts, and there are some corner cases in the Bourne shell that are not completely compatible with a POSIX shell. Thus, vendors who have taken this route will never (OK...“never say never”) replace the Bourne shell (as /bin/sh) with a POSIX shell.

Robert Lipe:

This is exactly the problem. While most (at least most System V's) do have a Bourne shell that accepts shell functions most vendor /bin/sh programs are not the POSIX shell.

So while most modern systems do have a shell somewhere that meets the POSIX standard, the challenge is to find it.

10.2 Here-Documents

Don't rely on \ being preserved just because it has no special meaning together with the next symbol. In the native /bin/sh on OpenBSD 2.7 \" expands to " in here-documents with unquoted delimiter. As a general rule, if \\ expands to \ use \\ to get \.

With OpenBSD 2.7's /bin/sh

     $ cat <<EOF
     > \" \\
     > EOF
     " \

and with Bash:

     bash-2.04$ cat <<EOF
     > \" \\
     > EOF
     \" \

Many older shells (including the Bourne shell) implement here-documents inefficiently. And some shells mishandle large here-documents: for example, Solaris 8 dtksh, which is derived from ksh M-12/28/93d, mishandles variable expansion that occurs on 1024-byte buffer boundaries within a here-document. Users can generally fix these problems by using a faster or more reliable shell, e.g., by using the command bash ./configure rather than plain ./configure.

Some shells can be extremely inefficient when there are a lot of here-documents inside a single statement. For instance if your includes something like:

     if <cross_compiling>; then
       assume this and that
       check this
       check that
       check something else
       on and on forever

A shell parses the whole if/fi construct, creating temporary files for each here document in it. Some shells create links for such here-documents on every fork, so that the clean-up code they had installed correctly removes them. It is creating the links that can take the shell forever.

Moving the tests out of the if/fi, or creating multiple if/fi constructs, would improve the performance significantly. Anyway, this kind of construct is not exactly the typical use of Autoconf. In fact, it's even not recommended, because M4 macros can't look into shell conditionals, so we may fail to expand a macro when it was expanded before in a conditional path, and the condition turned out to be false at run-time, and we end up not executing the macro at all.

10.3 File Descriptors

Some file descriptors shall not be used, since some systems, admittedly arcane, use them for special purpose:

     3 — some systems may open it to /dev/tty.
     4 — used on the Kubota Titan.

Don't redirect the same file descriptor several times, as you are doomed to failure under Ultrix.

     ULTRIX V4.4 (Rev. 69) System #31: Thu Aug 10 19:42:23 GMT 1995
     UWS V4.4 (Rev. 11)
     $ eval 'echo matter >fullness' >void
     illegal io
     $ eval '(echo matter >fullness)' >void
     illegal io
     $ (eval '(echo matter >fullness)') >void
     Ambiguous output redirect.

In each case the expected result is of course fullness containing matter and void being empty.

Don't try to redirect the standard error of a command substitution: it must be done inside the command substitution: when running : `cd /zorglub` 2>/dev/null expect the error message to escape, while : `cd /zorglub 2>/dev/null` works properly.

It is worth noting that Zsh (but not Ash nor Bash) makes it possible in assignments though: foo=`cd /zorglub` 2>/dev/null.

Most shells, if not all (including Bash, Zsh, Ash), output traces on stderr, even for sub-shells. This might result in undesirable content if you meant to capture the standard-error output of the inner command:

     $ ash -x -c '(eval "echo foo >&2") 2>stderr'
     $ cat stderr
     + eval echo foo >&2
     + echo foo
     $ bash -x -c '(eval "echo foo >&2") 2>stderr'
     $ cat stderr
     + eval 'echo foo >&2'
     ++ echo foo
     $ zsh -x -c '(eval "echo foo >&2") 2>stderr'
     # Traces on startup files deleted here.
     $ cat stderr
     +zsh:1> eval echo foo >&2
     +zsh:1> echo foo

You'll appreciate the various levels of detail....

One workaround is to grep out uninteresting lines, hoping not to remove good ones....

Don't try to move/delete open files, such as in exec >foo; mv foo bar; see Limitations of Builtins, mv for more details.

10.4 File System Conventions

While autoconf and friends will usually be run on some Unix variety, it can and will be used on other systems, most notably DOS variants. This impacts several assumptions regarding file and path names.

For example, the following code:

     case $foo_dir in
       /*) # Absolute
          foo_dir=$dots$foo_dir ;;

will fail to properly detect absolute paths on those systems, because they can use a drivespec, and will usually use a backslash as directory separator. The canonical way to check for absolute paths is:

     case $foo_dir in
       [\\/]* | ?:[\\/]* ) # Absolute
          foo_dir=$dots$foo_dir ;;

Make sure you quote the brackets if appropriate and keep the backslash as first character (see Limitations of Builtins).

Also, because the colon is used as part of a drivespec, these systems don't use it as path separator. When creating or accessing paths, use the PATH_SEPARATOR output variable instead. configure sets this to the appropriate value (: or ;) when it starts up.

File names need extra care as well. While DOS-based environments that are Unixy enough to run autoconf (such as DJGPP) will usually be able to handle long file names properly, there are still limitations that can seriously break packages. Several of these issues can be easily detected by the doschk package.

A short overview follows; problems are marked with sfn/lfn to indicate where they apply: sfn means the issues are only relevant to plain DOS, not to DOS boxes under Windows, while lfn identifies problems that exist even under Windows.

No multiple dots (sfn)
DOS cannot handle multiple dots in filenames. This is an especially important thing to remember when building a portable configure script, as autoconf uses a .in suffix for template files.

This is perfectly OK on Unices:


but it causes problems on DOS, as it requires, and To make your package more portable to DOS-based environments, you should use this instead:


No leading dot (sfn)
DOS cannot handle filenames that start with a dot. This is usually not a very important issue for autoconf.
Case insensitivity (lfn)
DOS is case insensitive, so you cannot, for example, have both a file called INSTALL and a directory called install. This also affects make; if there's a file called INSTALL in the directory, make install will do nothing (unless the install target is marked as PHONY).
The 8+3 limit (sfn)
Because the DOS file system only stores the first 8 characters of the filename and the first 3 of the extension, those must be unique. That means that foobar-part1.c, foobar-part2.c and foobar-prettybird.c all resolve to the same filename (FOOBAR-P.C). The same goes for and foo.bartender.

Note: This is not usually a problem under Windows, as it uses numeric tails in the short version of filenames to make them unique. However, a registry setting can turn this behavior off. While this makes it possible to share file trees containing long file names between sfn and lfn environments, it also means the above problem applies there as well.

Invalid characters
Some characters are invalid in DOS filenames, and should therefore be avoided. In a lfn environment, these are /, \, ?, *, :, <, >, | and ". In a sfn environment, other characters are also invalid. These include +, ,, [ and ].

10.5 Shell Substitutions

Contrary to a persistent urban legend, the Bourne shell does not systematically split variables and back-quoted expressions, in particular on the right-hand side of assignments and in the argument of case. For instance, the following code:

     case "$given_srcdir" in
     .)  top_srcdir="`echo "$dots" | sed 's,/$,,'`"
     *)  top_srcdir="$dots$given_srcdir" ;;

is more readable when written as:

     case $given_srcdir in
     .)  top_srcdir=`echo "$dots" | sed 's,/$,,'`
     *)  top_srcdir=$dots$given_srcdir ;;

and in fact it is even more portable: in the first case of the first attempt, the computation of top_srcdir is not portable, since not all shells properly understand "`..."..."...`". Worse yet, not all shells understand "`...\"...\"...`" the same way. There is just no portable way to use double-quoted strings inside double-quoted back-quoted expressions (pfew!).

One of the most famous shell-portability issues is related to "$@". When there are no positional arguments, POSIX says that "$@" is supposed to be equivalent to nothing, but the original Unix Version 7 Bourne shell treated it as equivalent to "" instead, and this behavior survives in later implementations like Digital Unix 5.0.

The traditional way to work around this portability problem is to use ${1+"$@"}. Unfortunately this method does not work with Zsh (3.x and 4.x), which is used on Mac OS X. When emulating the Bourne shell, Zsh performs word splitting on ${1+"$@"}:

          zsh $ emulate sh
          zsh $ for i in "$@"; do echo $i; done
          Hello World
          zsh $ for i in ${1+"$@"}; do echo $i; done

Zsh handles plain "$@" properly, but we can't use plain "$@" because of the portability problems mentioned above. One workaround relies on Zsh's “global aliases” to convert ${1+"$@"} into "$@" by itself:

          test "${ZSH_VERSION+set}" = set && alias -g '${1+"$@"}'='"$@"'

A more conservative workaround is to avoid "$@" if it is possible that there may be no positional arguments. For example, instead of:

          cat conftest.c "$@"

you can use this instead:

          case $# in
          0) cat conftest.c;;
          *) cat conftest.c "$@";;

Old BSD shells, including the Ultrix sh, don't accept the colon for any shell substitution, and complain and die.
Be sure to quote:
          : ${var='Some words'}

otherwise some shells, such as on Digital Unix V 5.0, will die because of a “bad substitution”.

Solaris' /bin/sh has a frightening bug in its interpretation of this. Imagine you need set a variable to a string containing }. This } character confuses Solaris' /bin/sh when the affected variable was already set. This bug can be exercised by running:
          $ unset foo
          $ foo=${foo='}'}
          $ echo $foo
          $ foo=${foo='}'   # no error; this hints to what the bug is
          $ echo $foo
          $ foo=${foo='}'}
          $ echo $foo
           ^ ugh!

It seems that } is interpreted as matching ${, even though it is enclosed in single quotes. The problem doesn't happen using double quotes.

On Ultrix, running
          : ${var="$default"}

will set var to M-yM-uM-,M-yM-aM-a, i.e., the 8th bit of each char will be set. You won't observe the phenomenon using a simple echo $var since apparently the shell resets the 8th bit when it expands $var. Here are two means to make this shell confess its sins:

          $ cat -v <<EOF


          $ set | grep '^var=' | cat -v

One classic incarnation of this bug is:

          default="a b c"
          : ${list="$default"}
          for c in $list; do
            echo $c

You'll get a b c on a single line. Why? Because there are no spaces in $list: there are M- , i.e., spaces with the 8th bit set, hence no IFS splitting is performed!!!

One piece of good news is that Ultrix works fine with : ${list=$default}; i.e., if you don't quote. The bad news is then that QNX 4.25 then sets list to the last item of default!

The portable way out consists in using a double assignment, to switch the 8th bit twice on Ultrix:


...but beware of the } bug from Solaris (see above). For safety, use:

          test "${var+set}" = set || var={value}

While in general it makes no sense, do not substitute a single builtin with side effects, becauase Ash 0.2, trying to optimize, does not fork a subshell to perform the command.

For instance, if you wanted to check that cd is silent, do not use test -z "`cd /`" because the following can happen:

          $ pwd
          $ test -z "`cd /`" && pwd

The result of foo=`exit 1` is left as an exercise to the reader.

This construct is meant to replace `commands`; they can be nested while this is impossible to do portably with back quotes. Unfortunately it is not yet widely supported. Most notably, even recent releases of Solaris don't support it:
          $ showrev -c /bin/sh | grep version
          Command version: SunOS 5.8 Generic 109324-02 February 2001
          $ echo $(echo blah)
          syntax error: `(' unexpected

nor does irix 6.5's Bourne shell:

          $ uname -a
          IRIX firebird-image 6.5 07151432 IP22
          $ echo $(echo blah)
          $(echo blah)

If you do use $(commands), make sure that the commands do not start with a parenthesis, as that would cause confusion with a different notation $((expression)) that in modern shells is an arithmetic expression not a command. To avoid the confusion, insert a space between the two opening parentheses.

10.6 Assignments

When setting several variables in a row, be aware that the order of the evaluation is undefined. For instance foo=1 foo=2; echo $foo gives 1 with sh on Solaris, but 2 with Bash. You must use ; to enforce the order: foo=1; foo=2; echo $foo.

Don't rely on the following to find subdir/program:

     PATH=subdir$PATH_SEPARATOR$PATH program

as this does not work with Zsh 3.0.6. Use something like this instead:

     (PATH=subdir$PATH_SEPARATOR$PATH; export PATH; exec program)

Don't rely on the exit status of an assignment: Ash 0.2 does not change the status and propagates that of the last statement:

     $ false || foo=bar; echo $?
     $ false || foo=`:`; echo $?

and to make things even worse, QNX 4.25 just sets the exit status to 0 in any case:

     $ foo=`exit 1`; echo $?

To assign default values, follow this algorithm:

  1. If the default value is a literal and does not contain any closing brace, use:
              : ${var='my literal'}
  2. If the default value contains no closing brace, has to be expanded, and the variable being initialized will never be IFS-split (i.e., it's not a list), then use:
              : ${var="$default"}
  3. If the default value contains no closing brace, has to be expanded, and the variable being initialized will be IFS-split (i.e., it's a list), then use:
  4. If the default value contains a closing brace, then use:
              test "${var+set}" = set || var='${indirection}'

In most cases var=${var="$default"} is fine, but in case of doubt, just use the latter. See Shell Substitutions, items ${var:-value} and ${var=value} for the rationale.

10.7 Parentheses in Shell Scripts

Beware of two opening parentheses in a row, as some shell implementations mishandle them. For example, pdksh 5.2.14 misparses the following code:

     if ((true) || false); then
       echo ok

To work around this problem, insert a space between the two opening parentheses. There is a similar problem and workaround with $((; see Shell Substitutions.

POSIX requires support for case patterns with opening parentheses like this:

     case $filename in
     (*.c) echo "C source code";;

but the ( in this example is not portable to many older Bourne shell implementations. It can be omitted safely.

10.8 Special Shell Variables

Some shell variables should not be used, since they can have a deep influence on the behavior of the shell. In order to recover a sane behavior from the shell, some variables should be unset, but unset is not portable (see Limitations of Builtins) and a fallback value is needed. We list these values below.

When this variable is set it specifies a list of directories to search when invoking cd with a relative filename. POSIX 1003.1-2001 says that if a nonempty directory name from CDPATH is used successfully, cd prints the resulting absolute filename. Unfortunately this output can break idioms like abs=`cd src && pwd` because abs receives the path twice. Also, many shells do not conform to this part of POSIX; for example, zsh prints the result only if a directory name other than . was chosen from CDPATH.

In practice the shells that have this problem also support unset, so you can work around the problem as follows:

          (unset CDPATH) >/dev/null 2>&1 && unset CDPATH

Autoconf-generated scripts automatically unset CDPATH if possible, so you need not worry about this problem in those scripts.

Don't set the first character of IFS to backslash. Indeed, Bourne shells use the first character (backslash) when joining the components in "$@" and some shells then re-interpret (!) the backslash escapes, so you can end up with backspace and other strange characters.

The proper value for IFS (in regular code, not when performing splits) is <SPC><TAB><RET>. The first character is especially important, as it is used to join the arguments in @*.

Autoconf-generated scripts normally set all these variables to C because so much configuration code assumes the C locale and POSIX requires that locale environment variables be set to C if the C locale is desired. However, some older, nonstandard systems (notably SCO) break if locale environment variables are set to C, so when running on these systems Autoconf-generated scripts unset the variables instead.
LANGUAGE is not specified by POSIX, but it is a GNU extension that overrides LC_ALL in some cases, so Autoconf-generated scripts set it too.
These locale environment variables are GNU extensions. They are treated like their POSIX brethren (LC_COLLATE, etc.) as described above.
Most modern shells provide the current line number in LINENO. Its value is the line number of the beginning of the current command. Autoconf attempts to execute configure with a modern shell. If no such shell is available, it attempts to implement LINENO with a Sed prepass that replaces each instance of the string $LINENO (not followed by an alphanumeric character) with the line's number.

You should not rely on LINENO within eval, as the behavior differs in practice. Also, the possibility of the Sed prepass means that you should not rely on $LINENO when quoted, when in here-documents, or when in long commands that cross line boundaries. Subshells should be OK, though. In the following example, lines 1, 6, and 9 are portable, but the other instances of LINENO are not:

          $ cat lineno
          echo 1. $LINENO
          cat <<EOF
          3. $LINENO
          4. $LINENO
          ( echo 6. $LINENO )
          eval 'echo 7. $LINENO'
          echo 8. '$LINENO'
          echo 9. $LINENO '
          10.' $LINENO
          $ bash-2.05 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 1
          8. $LINENO
          9. 9
          10. 9
          $ zsh-3.0.6 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 7
          8. $LINENO
          9. 9
          10. 9
          $ pdksh-5.2.14 lineno
          1. 1
          3. 2
          4. 2
          6. 6
          7. 0
          8. $LINENO
          9. 9
          10. 9
          $ sed '=' <lineno |
          >   sed '
          >     N
          >     s,$,-,
          >     : loop
          >     s,^\([0-9]*\)\(.*\)[$]LINENO\([^a-zA-Z0-9_]\),\1\2\1\3,
          >     t loop
          >     s,-$,,
          >     s,^[0-9]*\n,,
          >   ' |
          >   sh
          1. 1
          3. 3
          4. 4
          6. 6
          7. 7
          8. 8
          9. 9
          10. 10

When executing the command >foo, zsh executes $NULLCMD >foo. The Bourne shell considers NULLCMD to be :, while zsh, even in Bourne shell compatibility mode, sets NULLCMD to cat. If you forgot to set NULLCMD, your script might be suspended waiting for data on its standard input.
These variables should not matter for shell scripts, since they are supposed to affect only interactive shells. However, at least one shell (the pre-3.0 uwin ksh) gets confused about whether it is interactive, which means that (for example) a PS1 with a side effect can unexpectedly modify $?. To work around this bug, Autoconf-generated scripts do something like this:
          (unset ENV) >/dev/null 2>&1 && unset ENV MAIL MAILPATH
          PS1='$ '
          PS2='> '
          PS4='+ '

POSIX 1003.1-2001 requires that cd and pwd must update the PWD environment variable to point to the logical path to the current directory, but traditional shells do not support this. This can cause confusion if one shell instance maintains PWD but a subsidiary and different shell does not know about PWD and executes cd; in this case PWD will point to the wrong directory. Use `pwd` rather than $PWD.
This variable is an alias to $? for zsh (at least 3.1.6), hence read-only. Do not use it.
If it is not set, configure will detect the appropriate path separator for the build system and set the PATH_SEPARATOR output variable accordingly.

On DJGPP systems, the PATH_SEPARATOR environment variable can be set to either : or ; to control the path separator bash uses to set up certain environment variables (such as PATH). Since this only works inside bash, you want configure to detect the regular DOS path separator (;), so it can be safely substituted in files that may not support ; as path separator. So it is recommended to either unset this variable or set it to ;.

Many shells provide RANDOM, a variable that returns a different integer each time it is used. Most of the time, its value does not change when it is not used, but on irix 6.5 the value changes all the time. This can be observed by using set.

10.9 Limitations of Shell Builtins

No, no, we are serious: some shells do have limitations! :)

You should always keep in mind that any builtin or command may support options, and therefore have a very different behavior with arguments starting with a dash. For instance, the innocent echo "$word" can give unexpected results when word starts with a dash. It is often possible to avoid this problem using echo "x$word", taking the x into account later in the pipe.

Use . only with regular files (use test -f). Bash 2.03, for instance, chokes on . /dev/null. Also, remember that . uses PATH if its argument contains no slashes, so if you want to use . on a file foo in the current directory, you must use . ./foo.
You can't use !; you'll have to rewrite your code.
The use of break 2 etc. is safe.
POSIX 1003.1-2001 requires that cd must support the -L (“logical”) and -P (“physical”) options, with -L being the default. However, traditional shells do not support these options, and their cd command has the -P behavior.

Portable scripts should assume neither option is supported, and should assume neither behavior is the default. This can be a bit tricky, since the POSIX default behavior means that, for example, ls .. and cd .. may refer to different directories if the current logical directory is a symbolic link. It is safe to use cd dir if dir contains no .. components. Also, Autoconf-generated scripts check for this problem when computing variables like ac_top_srcdir (see Configuration Actions), so it is safe to cd to these variables.

Also please see the discussion of the pwd command.

You don't need to quote the argument; no splitting is performed.

You don't need the final ;;, but you should use it.

Because of a bug in its fnmatch, bash fails to properly handle backslashes in character classes:

          bash-2.02$ case /tmp in [/\\]*) echo OK;; esac

This is extremely unfortunate, since you are likely to use this code to handle unix or ms-dos absolute paths. To work around this bug, always put the backslash first:

          bash-2.02$ case '\TMP' in [\\/]*) echo OK;; esac
          bash-2.02$ case /tmp in [\\/]*) echo OK;; esac

Some shells, such as Ash 0.3.8, are confused by an empty case/esac:

          ash-0.3.8 $ case foo in esac;
          error-->Syntax error: ";" unexpected (expecting ")")

Many shells still do not support parenthesized cases, which is a pity for those of us using tools that rely on balanced parentheses. For instance, Solaris 8's Bourne shell:

          $ case foo in (foo) echo foo;; esac
          error-->syntax error: `(' unexpected

The simple echo is probably the most surprising source of portability troubles. It is not possible to use echo portably unless both options and escape sequences are omitted. New applications which are not aiming at portability should use printf instead of echo.

Don't expect any option. See Preset Output Variables, ECHO_N etc. for a means to simulate -n.

Do not use backslashes in the arguments, as there is no consensus on their handling. On echo '\n' | wc -l, the sh of Digital Unix 4.0 and MIPS RISC/OS 4.52, answer 2, but the Solaris' sh, Bash, and Zsh (in sh emulation mode) report 1. Please note that the problem is truly echo: all the shells understand '\n' as the string composed of a backslash and an n.

Because of these problems, do not pass a string containing arbitrary characters to echo. For example, echo "$foo" is safe if you know that foo's value cannot contain backslashes and cannot start with -, but otherwise you should use a here-document like this:

          cat <<EOF

The default value of exit is supposed to be $?; unfortunately, some shells, such as the DJGPP port of Bash 2.04, just perform exit 0.
          bash-2.04$ foo=`exit 1` || echo fail
          bash-2.04$ foo=`(exit 1)` || echo fail
          bash-2.04$ foo=`(exit 1); exit` || echo fail

Using exit $? restores the expected behavior.

Some shell scripts, such as those generated by autoconf, use a trap to clean up before exiting. If the last shell command exited with nonzero status, the trap also exits with nonzero status so that the invoker can tell that an error occurred.

Unfortunately, in some shells, such as Solaris 8 sh, an exit trap ignores the exit command's argument. In these shells, a trap cannot determine whether it was invoked by plain exit or by exit 1. Instead of calling exit directly, use the AC_MSG_ERROR macro that has a workaround for this problem.

The builtin export dubs a shell variable environment variable. Each update of exported variables corresponds to an update of the environment variables. Conversely, each environment variable received by the shell when it is launched should be imported as a shell variable marked as exported.

Alas, many shells, such as Solaris 2.5, irix 6.3, irix 5.2, AIX 4.1.5, and Digital unix 4.0, forget to export the environment variables they receive. As a result, two variables coexist: the environment variable and the shell variable. The following code demonstrates this failure:

          #! /bin/sh
          echo $FOO
          echo $FOO
          exec /bin/sh $0

when run with FOO=foo in the environment, these shells will print alternately foo and bar, although it should only print foo and then a sequence of bars.

Therefore you should export again each environment variable that you update.

Don't expect false to exit with status 1: in the native Bourne shell of Solaris 8 it exits with status 255.
To loop over positional arguments, use:
          for arg
            echo "$arg"

You may not leave the do on the same line as for, since some shells improperly grok:

          for arg; do
            echo "$arg"

If you want to explicitly refer to the positional arguments, given the $@ bug (see Shell Substitutions), use:

          for arg in ${1+"$@"}; do
            echo "$arg"

But keep in mind that Zsh, even in Bourne shell emulation mode, performs word splitting on ${1+"$@"}; see Shell Substitutions, item $@, for more.

Using ! is not portable. Instead of:
          if ! cmp -s file; then
            mv file


          if cmp -s file; then :; else
            mv file

There are shells that do not reset the exit status from an if:

          $ if (exit 42); then true; fi; echo $?

whereas a proper shell should have printed 0. This is especially bad in Makefiles since it produces false failures. This is why properly written Makefiles, such as Automake's, have such hairy constructs:

          if test -f "$file"; then
            install "$file" "$dest"

A format string starting with a - can cause problems. bash (eg. 2.05b) will interpret it as an options string and give an error. And -- to mark the end of options is not good in the NetBSD Almquist shell (eg. 0.4.6) which will take that literally as the format string. Putting the - in a %c or %s is probably the easiest way to avoid doubt,
          printf %s -foo

With modern shells, plain pwd outputs a “logical” directory name, some of whose components may be symbolic links. These directory names are in contrast to “physical” directory names, whose components are all directories.

POSIX 1003.1-2001 requires that pwd must support the -L (“logical”) and -P (“physical”) options, with -L being the default. However, traditional shells do not support these options, and their pwd command has the -P behavior.

Portable scripts should assume neither option is supported, and should assume neither behavior is the default. Also, on many hosts /bin/pwd is equivalent to pwd -P, but POSIX does not require this behavior and portable scripts should not rely on it.

Typically it's best to use plain pwd. On modern hosts this outputs logical directory names, which have the following advantages:

Also please see the discussion of the cd command.

This builtin faces the usual problem with arguments starting with a dash. Modern shells such as Bash or Zsh understand -- to specify the end of the options (any argument after -- is a parameter, even -x for instance), but most shells simply stop the option processing as soon as a non-option argument is found. Therefore, use dummy or simply x to end the option processing, and use shift to pop it out:
          set x $my_list; shift

Some shells have the "opposite" problem of not recognizing all options (e.g., set -e -x assigns -x to the command line). It is better to elide these:

          set -ex

Not only is shifting a bad idea when there is nothing left to shift, but in addition it is not portable: the shell of MIPS RISC/OS 4.52 refuses to do it.
This command is not portable, as POSIX does not require it; use . instead.
The test program is the way to perform many file and string tests. It is often invoked by the alternate name [, but using that name in Autoconf code is asking for trouble since it is an M4 quote character.

If you need to make multiple checks using test, combine them with the shell operators && and || instead of using the test operators -a and -o. On System V, the precedence of -a and -o is wrong relative to the unary operators; consequently, POSIX does not specify them, so using them is nonportable. If you combine && and || in the same statement, keep in mind that they have equal precedence.

You may use ! with test, but not with if: test ! -r foo || exit 1.

test (files)
To enable configure scripts to support cross-compilation, they shouldn't do anything that tests features of the build system instead of the host system. But occasionally you may find it necessary to check whether some arbitrary file exists. To do so, use test -f or test -r. Do not use test -x, because 4.3BSD does not have it. Do not use test -e either, because Solaris 2.5 does not have it. To test for symbolic links on systems that have them, use test -h rather than test -L; either form conforms to POSIX 1003.1-2001, but older shells like Solaris 8 /bin/sh support only -h.
test (strings)
Avoid test "string", in particular if string might start with a dash, since test might interpret its argument as an option (e.g., string = "-n").

Contrary to a common belief, test -n string and test -z string are portable. Nevertheless many shells (such as Solaris 2.5, AIX 3.2, unicos, Digital Unix 4 etc.) have bizarre precedence and may be confused if string looks like an operator:

          $ test -n =
          test: argument expected

If there are risks, use test "xstring" = x or test "xstring" != x instead.

It is common to find variations of the following idiom:

          test -n "`echo $ac_feature | sed 's/[-a-zA-Z0-9_]//g'`" &&

to take an action when a token matches a given pattern. Such constructs should always be avoided by using:

          echo "$ac_feature" | grep '[^-a-zA-Z0-9_]' >/dev/null 2>&1 &&

Use case where possible since it is faster, being a shell builtin:

          case $ac_feature in
            *[!-a-zA-Z0-9_]*) action;;

Alas, negated character classes are probably not portable, although no shell is known to not support the POSIX syntax [!...] (when in interactive mode, zsh is confused by the [!...] syntax and looks for an event in its history because of !). Many shells do not support the alternative syntax [^...] (Solaris, Digital Unix, etc.).

One solution can be:

          expr "$ac_feature" : '.*[^-a-zA-Z0-9_]' >/dev/null &&

or better yet

          expr "x$ac_feature" : '.*[^-a-zA-Z0-9_]' >/dev/null &&

expr "Xfoo" : "Xbar" is more robust than echo "Xfoo" | grep "^Xbar", because it avoids problems when foo contains backslashes.

It is safe to trap at least the signals 1, 2, 13, and 15. You can also trap 0, i.e., have the trap run when the script ends (either via an explicit exit, or the end of the script).

Although POSIX is not absolutely clear on this point, it is widely admitted that when entering the trap $? should be set to the exit status of the last command run before the trap. The ambiguity can be summarized as: “when the trap is launched by an exit, what is the last command run: that before exit, or exit itself?”

Bash considers exit to be the last command, while Zsh and Solaris 8 sh consider that when the trap is run it is still in the exit, hence it is the previous exit status that the trap receives:

          $ cat
          trap 'echo $?' 0
          (exit 42); exit 0
          $ zsh
          $ bash

The portable solution is then simple: when you want to exit 42, run (exit 42); exit 42, the first exit being used to set the exit status to 42 for Zsh, and the second to trigger the trap and pass 42 as exit status for Bash.

The shell in FreeBSD 4.0 has the following bug: $? is reset to 0 by empty lines if the code is inside trap.

          $ trap 'false
          echo $?' 0
          $ exit

Fortunately, this bug only affects trap.

Don't worry: as far as we know true is portable. Nevertheless, it's not always a builtin (e.g., Bash 1.x), and the portable shell community tends to prefer using :. This has a funny side effect: when asked whether false is more portable than true Alexandre Oliva answered:
In a sense, yes, because if it doesn't exist, the shell will produce an exit status of failure, which is correct for false, but not for true.

You cannot assume the support of unset. Nevertheless, because it is extremely useful to disable embarrassing variables such as PS1, you can test for its existence and use it provided you give a neutralizing value when unset is not supported:
          if (unset FOO) >/dev/null 2>&1; then
          $unset PS1 || PS1='$ '

See Special Shell Variables, for some neutralizing values. Also, see Limitations of Builtins, documentation of export, for the case of environment variables.

10.10 Limitations of Usual Tools

The small set of tools you can expect to find on any machine can still include some limitations you should be aware of.

Don't leave white spaces before the parentheses in user functions calls; GNU awk will reject it:
          $ gawk 'function die () { print "Aaaaarg!"  }
                  BEGIN { die () }'
          gawk: cmd. line:2:         BEGIN { die () }
          gawk: cmd. line:2:                      ^ parse error
          $ gawk 'function die () { print "Aaaaarg!"  }
                  BEGIN { die() }'

If you want your program to be deterministic, don't depend on for on arrays:

          $ cat for.awk
          END {
            arr["foo"] = 1
            arr["bar"] = 1
            for (i in arr)
              print i
          $ gawk -f for.awk </dev/null
          $ nawk -f for.awk </dev/null

Some AWK, such as HPUX 11.0's native one, have regex engines fragile to inner anchors:

          $ echo xfoo | $AWK '/foo|^bar/ { print }'
          $ echo bar | $AWK '/foo|^bar/ { print }'
          $ echo xfoo | $AWK '/^bar|foo/ { print }'
          $ echo bar | $AWK '/^bar|foo/ { print }'

Either do not depend on such patterns (i.e., use /^(.*foo|bar)/, or use a simple test to reject such AWK.

Don't rely on any option. The option -v, which displays non-printing characters, seems portable, though.
When a compilation such as cc foo.c -o foo fails, some compilers (such as cds on Reliant unix) leave a foo.o.

HP-UX cc doesn't accept .S files to preprocess and assemble. cc -c foo.S will appear to succeed, but in fact does nothing.

The default executable, produced by cc foo.c, can be

cmp performs a raw data comparison of two files, while diff compares two text files. Therefore, if you might compare DOS files, even if only checking whether two files are different, use diff to avoid spurious differences due to differences of newline encoding.
Traditionally, file timestamps had 1-second resolution, and cp -p copied the timestamps exactly. However, many modern filesystems have timestamps with 1-nanosecond resolution. Unfortunately, cp -p implementations truncate timestamps when copying files, so this can result in the destination file appearing to be older than the source. The exact amount of truncation depends on the resolution of the system calls that cp uses; traditionally this was utime, which has 1-second resolution, but some newer cp implementations use utimes, which has 1-microsecond resolution. These newer implementations include GNU coreutils 5.0.91 or later, and Solaris 8 (sparc) patch 109933-02 or later. Unfortunately as of September 2003 there is still no system call to set time stamps to the full nanosecond resolution.

SunOS cp does not support -f, although its mv does. It's possible to deduce why mv and cp are different with respect to -f. mv prompts by default before overwriting a read-only file. cp does not. Therefore, mv requires a -f option, but cp does not. mv and cp behave differently with respect to read-only files because the simplest form of cp cannot overwrite a read-only file, but the simplest form of mv can. This is because cp opens the target for write access, whereas mv simply calls link (or, in newer systems, rename).

Bob Proulx notes that cp -p always tries to copy ownerships. But whether it actually does copy ownerships or not is a system dependent policy decision implemented by the kernel. If the kernel allows it then it happens. If the kernel does not allow it then it does not happen. It is not something cp itself has control over.

In SysV any user can chown files to any other user, and SysV also had a non-sticky /tmp. That undoubtedly derives from the heritage of SysV in a business environment without hostile users. BSD changed this to be a more secure model where only root can chown files and a sticky /tmp is used. That undoubtedly derives from the heritage of BSD in a campus environment.

Linux by default follows BSD, but it can be configured to allow chown. HP-UX as an alternate example follows SysV, but it can be configured to use the modern security model and disallow chown. Since it is an administrator configurable parameter you can't use the name of the kernel as an indicator of the behavior.

Some versions of date do not recognize special % directives, and unfortunately, instead of complaining, they just pass them through, and exit with success:
          $ uname -a
          OSF1 V5.1 732 alpha
          $ date "+%s"

Option -u is nonportable.

Some implementations, such as Tru64's, fail when comparing to /dev/null. Use an empty file instead.

Not all hosts have a working dirname, and you should instead use AS_DIRNAME (see Programming in M4sh). For example:
          dir=`dirname "$file"`       # This is not portable.
          dir=`AS_DIRNAME(["$file"])` # This is more portable.

This handles a few subtleties in the standard way required by POSIX. For example, under UN*X, should dirname //1 give /? Paul Eggert answers:

No, under some older flavors of Unix, leading // is a special path name: it refers to a “super-root” and is used to access other machines' files. Leading ///, ////, etc. are equivalent to /; but leading // is special. I think this tradition started with Apollo Domain/OS, an OS that is still in use on some older hosts.

POSIX allows but does not require the special treatment for //. It says that the behavior of dirname on path names of the form //([^/]+/*)? is implementation defined. In these cases, GNU dirname returns /, but it's more portable to return // as this works even on those older flavors of Unix.

POSIX 1003.1-2001 no longer requires egrep, but many older hosts do not yet support the POSIX replacement grep -E. To work around this problem, invoke AC_PROG_EGREP and then use $EGREP.

The empty alternative is not portable, use ? instead. For instance with Digital Unix v5.0:

          > printf "foo\n|foo\n" | $EGREP '^(|foo|bar)$'
          > printf "bar\nbar|\n" | $EGREP '^(foo|bar|)$'
          > printf "foo\nfoo|\n|bar\nbar\n" | $EGREP '^(foo||bar)$'

$EGREP also suffers the limitations of grep.

No expr keyword starts with x, so use expr x"word" : 'xregex' to keep expr from misinterpreting word.

Don't use length, substr, match and index.

expr (|)
You can use |. Although POSIX does require that expr '' return the empty string, it does not specify the result when you | together the empty string (or zero) with the empty string. For example:
          expr '' \| ''

GNU/Linux and POSIX.2-1992 return the empty string for this case, but traditional unix returns 0 (Solaris is one such example). In POSIX.1-2001, the specification has been changed to match traditional unix's behavior (which is bizarre, but it's too late to fix this). Please note that the same problem does arise when the empty string results from a computation, as in:

          expr bar : foo \| foo : bar

Avoid this portability problem by avoiding the empty string.

expr (:)
Don't use \?, \+ and \| in patterns, as they are not supported on Solaris.

The POSIX standard is ambiguous as to whether expr 'a' : '\(b\)' outputs 0 or the empty string. In practice, it outputs the empty string on most platforms, but portable scripts should not assume this. For instance, the QNX 4.25 native expr returns 0.

One might think that a way to get a uniform behavior would be to use the empty string as a default value:

          expr a : '\(b\)' \| ''

Unfortunately this behaves exactly as the original expression; see the expr (:) entry for more information.

Older expr implementations (e.g., SunOS 4 expr and Solaris 8 /usr/ucb/expr) have a silly length limit that causes expr to fail if the matched substring is longer than 120 bytes. In this case, you might want to fall back on echo|sed if expr fails.

Don't leave, there is some more!

The QNX 4.25 expr, in addition of preferring 0 to the empty string, has a funny behavior in its exit status: it's always 1 when parentheses are used!

          $ val=`expr 'a' : 'a'`; echo "$?: $val"
          0: 1
          $ val=`expr 'a' : 'b'`; echo "$?: $val"
          1: 0
          $ val=`expr 'a' : '\(a\)'`; echo "?: $val"
          1: a
          $ val=`expr 'a' : '\(b\)'`; echo "?: $val"
          1: 0

In practice this can be a big problem if you are ready to catch failures of expr programs with some other method (such as using sed), since you may get twice the result. For instance

          $ expr 'a' : '\(a\)' || echo 'a' | sed 's/^\(a\)$/\1/'

will output a on most hosts, but aa on QNX 4.25. A simple workaround consists in testing expr and use a variable set to expr or to false according to the result.

POSIX 1003.1-2001 no longer requires fgrep, but many older hosts do not yet support the POSIX replacement grep -F. To work around this problem, invoke AC_PROG_FGREP and then use $FGREP.
The option -maxdepth seems to be GNU specific. Tru64 v5.1, NetBSD 1.5 and Solaris 2.5 find commands do not understand it.

The replacement of {} is guaranteed only if the argument is exactly {}, not if it's only a part of an argument. For instance on DU, and HP-UX 10.20 and HP-UX 11:

          $ touch foo
          $ find . -name foo -exec echo "{}-{}" \;

while GNU find reports ./foo-./foo.

Don't use grep -s to suppress output, because grep -s on System V does not suppress output, only error messages. Instead, redirect the standard output and standard error (in case the file doesn't exist) of grep to /dev/null. Check the exit status of grep to determine whether it found a match.

Don't use multiple regexps with -e, as some grep will only honor the last pattern (e.g., irix 6.5 and Solaris 2.5.1). Anyway, Stardent Vistra SVR4 grep lacks -e... Instead, use extended regular expressions and alternation.

Don't rely on -w, as Irix 6.5.16m's grep does not support it.

Don't rely on ln having a -f option. Symbolic links are not available on old systems; use $(LN_S) as a portable substitute.

For versions of the DJGPP before 2.04, ln emulates soft links to executables by generating a stub that in turn calls the real program. This feature also works with nonexistent files like in the Unix spec. So ln -s file link will generate link.exe, which will attempt to call file.exe if run. But this feature only works for executables, so cp -p is used instead for these systems. DJGPP versions 2.04 and later have full symlink support.

The portable options are -acdilrtu. Modern practice is for -l to output both owner and group, but traditional ls omits the group.

Modern practice is for all diagnostics to go to standard error, but traditional ls foo prints the message foo not found to standard output if foo does not exist. Be careful when writing shell commands like sources=`ls *.c 2>/dev/null`, since with traditional ls this is equivalent to sources="*.c not found" if there are no .c files.

None of mkdir's options are portable. Instead of mkdir -p filename, you should use use AS_MKDIR_P(filename) (see Programming in M4sh).
The only portable options are -f and -i.

Moving individual files between file systems is portable (it was in V6), but it is not always atomic: when doing mv new existing, there's a critical section where neither the old nor the new version of existing actually exists.

Be aware that moving files from /tmp can sometimes cause undesirable (but perfectly valid) warnings, even if you created these files. On some systems, creating the file in /tmp is setting a guid wheel which you may not be part of. So the file is copied, and then the chgrp fails:

          $ touch /tmp/foo
          $ mv /tmp/foo .
          error-->mv: ./foo: set owner/group (was: 3830/0): Operation not permitted
          $ echo $?
          $ ls foo

This behavior conforms to POSIX:

If the duplication of the file characteristics fails for any reason, mv shall write a diagnostic message to standard error, but this failure shall not cause mv to modify its exit status.”

Moving directories across mount points is not portable, use cp and rm.

Moving/Deleting open files isn't portable. The following can't be done on DOS/WIN32:

          exec > foo
          mv foo bar

nor can

          exec > foo
          rm -f foo

Patterns should not include the separator (unless escaped), even as part of a character class. In conformance with POSIX, the Cray sed will reject s/[^/]*$//: use s,[^/]*$,,.

Sed scripts should not use branch labels longer than 8 characters and should not contain comments.

Don't include extra ;, as some sed, such as NetBSD 1.4.2's, try to interpret the second as a command:

          $ echo a | sed 's/x/x/;;s/x/x/'
          sed: 1: "s/x/x/;;s/x/x/": invalid command code ;

Input should have reasonably long lines, since some sed have an input buffer limited to 4000 bytes.

Alternation, \|, is common but POSIX does not require its support, so it should be avoided in portable scripts. Solaris 8 sed does not support alternation; e.g., sed '/a\|b/d' deletes only lines that contain the literal string a|b.

Anchors (^ and $) inside groups are not portable.

Nested parenthesization in patterns (e.g., \(\(a*\)b*)\)) is quite portable to modern hosts, but is not supported by some older sed implementations like SVR3.

Of course the option -e is portable, but it is not needed. No valid Sed program can start with a dash, so it does not help disambiguating. Its sole usefulness is to help enforcing indentation as in:

          sed -e instruction-1 \
              -e instruction-2

as opposed to

          sed instruction-1;instruction-2

Contrary to yet another urban legend, you may portably use & in the replacement part of the s command to mean “what was matched”. All descendants of Bell Lab's V7 sed (at least; we don't have first hand experience with older seds) have supported it.

POSIX requires that you must not have any white space between ! and the following command. It is OK to have blanks between the address and the !. For instance, on Solaris 8:

          $ echo "foo" | sed -n '/bar/ ! p'
          error-->Unrecognized command: /bar/ ! p
          $ echo "foo" | sed -n '/bar/! p'
          error-->Unrecognized command: /bar/! p
          $ echo "foo" | sed -n '/bar/ !p'

sed (t)
Some old systems have sed that “forget” to reset their t flag when starting a new cycle. For instance on MIPS RISC/OS, and on irix 5.3, if you run the following sed script (the line numbers are not actual part of the texts):
          s/keep me/kept/g  # a
          t end             # b
          s/.*/deleted/g    # c
          : end             # d


          delete me         # 1
          delete me         # 2
          keep me           # 3
          delete me         # 4

you get

          delete me

instead of


Why? When processing 1, a matches, therefore sets the t flag, b jumps to d, and the output is produced. When processing line 2, the t flag is still set (this is the bug). Line a fails to match, but sed is not supposed to clear the t flag when a substitution fails. Line b sees that the flag is set, therefore it clears it, and jumps to d, hence you get delete me instead of deleted. When processing 3, t is clear, a matches, so the flag is set, hence b clears the flags and jumps. Finally, since the flag is clear, 4 is processed properly.

There are two things one should remember about t in sed. Firstly, always remember that t jumps if some substitution succeeded, not only the immediately preceding substitution. Therefore, always use a fake t clear; : clear to reset the t flag where indeed.

Secondly, you cannot rely on sed to clear the flag at each new cycle.

One portable implementation of the script above is:

          t clear
          : clear
          s/keep me/kept/g
          t end
          : end

If you specify the desired timestamp (e.g., with the -r option), touch typically uses the utime or utimes system call, which can result in the same kind of timestamp truncation problems that cp -p has.

On some old BSD systems, touch or any command that results in an empty file does not update the timestamps, so use a command like echo as a workaround.

GNU touch 3.16r (and presumably all before that) fails to work on SunOS 4.1.3 when the empty file is on an NFS-mounted 4.2 volume.

10.11 Limitations of Make

make itself suffers a great number of limitations, only a few of which are listed here. First of all, remember that since commands are executed by the shell, all its weaknesses are inherited....

POSIX says that the $< construct in makefiles can be used only in inference rules and in the .DEFAULT rule; its meaning in ordinary rules is unspecified. Solaris 8's make for instance will replace it with the argument.
Leading underscore in macro names
Some makes don't support leading underscores in macro names, such as on NEWS-OS 4.2R.
          $ cat Makefile
          _am_include = #
          _am_quote =
          all:; @echo this is test
          $ make
          Make: Must be a separator on rules line 2.  Stop.
          $ cat Makefile2
          am_include = #
          am_quote =
          all:; @echo this is test
          $ make -f Makefile2
          this is test

Trailing backslash in macro
On some versions of HP-UX, make will read multiple newlines following a backslash, continuing to the next non-empty line. For example,
          FOO = one \
          BAR = two
                  : FOO is "$(FOO)"
                  : BAR is "$(BAR)"

shows FOO equal to one BAR = two. Other makes sensibly let a backslash continue only to the immediately following line.

Escaped newline in comments
According to POSIX, Makefile comments start with # and continue until an unescaped newline is reached.
          % cat Makefile
          # A = foo \
                bar \
                  @echo ok
          % make   # GNU make

However in Real World this is not always the case. Some implementations discards anything from # up to the end of line, ignoring any trailing backslash.

          % pmake  # BSD make
          "Makefile", line 3: Need an operator
          Fatal errors encountered -- cannot continue

Therefore, if you want to comment out a multi-line definition, prefix each line with #, not only the first.

          # A = foo \
          #     bar \
          #     baz

make macro=value and sub-makes.
A command-line variable definition such as foo=bar overrides any definition of foo in the Makefile. Some make implementations (such as GNU make) will propagate this override to sub-invocations of make. Some other implementation will not pass the substitution along to sub-makes.
          % cat Makefile
          foo = foo
                  @echo $(foo)
                  $(MAKE) two
                  @echo $(foo)
          % make foo=bar            # GNU make 3.79.1
          make two
          make[1]: Entering directory `/home/adl'
          make[1]: Leaving directory `/home/adl'
          % pmake foo=bar           # BSD make
          pmake two

You have a few possibilities if you do want the foo=bar override to propagate to sub-makes. One is to use the -e option, which causes all environment variables to have precedence over the Makefile macro definitions, and declare foo as an environment variable:

          % env foo=bar make -e

The -e option is propagated to sub-makes automatically, and since the environment is inherited between make invocations, the foo macro will be overridden in sub-makes as expected.

This syntax (foo=bar make -e) is portable only when used outside a Makefile, for instance from a script or from the command line. When run inside a make rule, GNU make 3.80 and prior versions forget to propagate the -e option to sub-makes.

Moreover, using -e could have unexpected side-effects if your environment contains some other macros usually defined by the Makefile. (See also the note about make -e and SHELL below.)

Another way to propagate overrides to sub-makes is to do it manually, from your Makefile:

          foo = foo
                  @echo $(foo)
                  $(MAKE) foo=$(foo) two
                  @echo $(foo)

You need to foresee all macros that a user might want to override if you do that.

The SHELL macro
POSIX-compliant makes internally use the $(SHELL) macro to spawn shell processes and execute Makefile rules. This is a builtin macro supplied by make, but it can be modified from the Makefile or a command-line argument.

Not all makes will define this SHELL macro. OSF/Tru64 make is an example; this implementation will always use /bin/sh. So it's a good idea to always define SHELL in your Makefiles. If you use Autoconf, do

          SHELL = @SHELL@

POSIX-compliant makes should never acquire the value of $(SHELL) from the environment, even when make -e is used (otherwise, think about what would happen to your rules if SHELL=/bin/tcsh).

However not all make implementations will make this exception. For instance it's not surprising that OSF/Tru64 make doesn't protect SHELL, since it doesn't use it.

          % cat Makefile
          SHELL = /bin/sh
          FOO = foo
                  @echo $(SHELL)
                  @echo $(FOO)
          % env SHELL=/bin/tcsh FOO=bar make -e   # OSF1 V4.0 Make
          % env SHELL=/bin/tcsh FOO=bar gmake -e  # GNU make

Comments in rules
Never put comments in a rule.

Some make treat anything starting with a tab as a command for the current rule, even if the tab is immediately followed by a #. The make from Tru64 Unix V5.1 is one of them. The following Makefile will run # foo through the shell.

                  # foo

The obj/ subdirectory.
Never name one of your subdirectories obj/ if you don't like surprises.

If an obj/ directory exists, BSD make will enter it before reading Makefile. Hence the Makefile in the current directory will not be read.

          % cat Makefile
                  echo Hello
          % cat obj/Makefile
                  echo World
          % make      # GNU make
          echo Hello
          % pmake     # BSD make
          echo World

make -k
Do not rely on the exit status of make -k. Some implementations reflect whether they encountered an error in their exit status; other implementations always succeed.
          % cat Makefile
          % make -k; echo exit status: $?    # GNU make
          make: *** [all] Error 1
          exit status: 2
          % pmake -k; echo exit status: $?   # BSD make
          *** Error code 1 (continuing)
          exit status: 0

There is no VPATH support specified in POSIX. Many makes have a form of VPATH support, but its implementation is not consistent amongst makes.

Maybe the best suggestion to give to people who need the VPATH feature is to choose a make implementation and stick to it. Since the resulting Makefiles are not portable anyway, better choose a portable make (hint, hint).

Here are a couple of known issues with some VPATH implementations.

VPATH and double-colon rules
Any assignment to VPATH causes Sun make to only execute the first set of double-colon rules. (This comment has been here since 1994 and the context has been lost. It's probably about SunOS 4. If you can reproduce this, please send us a test case for illustration.)
$< not supported in explicit rules
As said elsewhere, using $< in explicit rules is not portable. The prerequisite file must be named explicitly in the rule. If you want to find the prerequisite via a VPATH search, you have to code the whole thing manually. For instance, using the following pattern:
               VPATH = ../src
               foo.o: foo.c
                       cc -c `test -f foo.c || echo ../src/`foo.c -o foo.o

Automatic rule rewriting
Some make implementations, such as SunOS make, will search prerequisites in VPATH and rewrite all their occurrences in the rule appropriately.

For instance

               VPATH = ../src
               foo.o: foo.c
                       cc -c foo.c -o foo.o

would execute cc -c ../src/foo.c -o foo.o if foo.c was found in ../src. That sounds great.

However, for the sake of other make implementations, we can't rely on this, and we have to search VPATH manually:

               VPATH = ../src
               foo.o: foo.c
                       cc -c `test -f foo.c || echo ../src/`foo.c -o foo.o

However the "prerequisite rewriting" still applies here. So if foo.c is in ../src, SunOS make will execute

               cc -c `test -f ../src/foo.c || echo ../src/`foo.c -o foo.o

which reduces to

               cc -c foo.c -o foo.o

and thus fails. Oops.

One workaround is to make sure that foo.c never appears as a plain word in the rule. For instance these three rules would be safe.

               VPATH = ../src
               foo.o: foo.c
                       cc -c `test -f ./foo.c || echo ../src/`foo.c -o foo.o
               foo2.o: foo2.c
                       cc -c `test -f 'foo2.c' || echo ../src/`foo2.c -o foo2.o
               foo3.o: foo3.c
                       cc -c `test -f "foo3.c" || echo ../src/`foo3.c -o foo3.o

Things get worse when your prerequisites are in a macro.

               VPATH = ../src
               HEADERS = foo.h foo2.h foo3.h
               install-HEADERS: $(HEADERS)
                       for i in $(HEADERS); do \
                         $(INSTALL) -m 644 `test -f $$i || echo ../src/`$$i \
                           $(DESTDIR)$(includedir)/$$i; \

The above install-HEADERS rule is not SunOS-proof because for i in $(HEADERS); will be expanded as for i in foo.h foo2.h foo3.h; where foo.h and foo2.h are plain words and are hence subject to VPATH adjustments.

If the three files are in ../src, the rule is run as:

               for i in ../src/foo.h ../src/foo2.h foo3.h; do \
                 install -m 644 `test -f $i || echo ../src/`$i \
                    /usr/local/include/$i; \

where the two first install calls will fail. For instance, consider the foo.h installation:

               install -m 644 `test -f ../src/foo.h || echo ../src/`../src/foo.h \

It reduces to:

               install -m 644 ../src/foo.h /usr/local/include/../src/foo.h;

Note that the manual VPATH search did not cause any problems here; however this command installs foo.h in an incorrect directory.

Trying to quote $(HEADERS) in some way, as we did for foo.c a few Makefiles ago, does not help:

               install-HEADERS: $(HEADERS)
                       headers='$(HEADERS)'; for i in $$headers; do \
                         $(INSTALL) -m 644 `test -f $$i || echo ../src/`$$i \
                           $(DESTDIR)$(includedir)/$$i; \

Indeed, headers='$(HEADERS)' expands to headers='foo.h foo2.h foo3.h' where foo2.h is still a plain word. (Aside: the headers='$(HEADERS)'; for i in $$headers; idiom is a good idea if $(HEADERS) can be empty, because some shell produce a syntax error on for i in;.)

One workaround is to strip this unwanted ../src/ prefix manually:

               VPATH = ../src
               HEADERS = foo.h foo2.h foo3.h
               install-HEADERS: $(HEADERS)
                       headers='$(HEADERS)'; for i in $$headers; do \
                         i=`expr "$$i" : '../src/\(.*\)'`;
                         $(INSTALL) -m 644 `test -f $$i || echo ../src/`$$i \
                           $(DESTDIR)$(includedir)/$$i; \

Automake does something similar.

OSF/Tru64 make creates prerequisite directories magically
When a prerequisite is a sub-directory of VPATH, Tru64 make will create it in the current directory.
               % mkdir -p foo/bar build
               % cd build
               % cat >Makefile <<END
               VPATH = ..
               all: foo/bar
               % make
               mkdir foo
               mkdir foo/bar

This can yield unexpected results if a rule uses a manual VPATH search as presented before.

               VPATH = ..
               all : foo/bar
                       command `test -d foo/bar || echo ../`foo/bar

The above command will be run on the empty foo/bar directory that was created in the current directory.

target lookup
GNU make uses a rather complex algorithm to decide when it should use files found via a VPATH search. See How Directory Searches are Performed (The GNU Make Manual).

If a target needs to be rebuilt, GNU make discards the filename found during the VPATH search for this target, and builds the file locally using the filename given in the Makefile. If a target does not need to be rebuilt, GNU make uses the filename found during the VPATH search.

Other make implementations, like NetBSD make, are easier to describe: the filename found during the VPATH search will be used whether the target needs to be rebuilt or not. Therefore new files are created locally, but existing files are updated at their VPATH location.

OpenBSD and FreeBSD makes, however, will never perform a VPATH search for a dependency which has an explicit rule. This is extremely annoying.

When attempting a VPATH build for an autoconfiscated package (e.g,, mkdir build && cd build && ../configure), this means the GNU make will build everything locally in the build directory, while BSD make will build new files locally and update existing files in the source directory.

               % cat Makefile
               VPATH = ..
               all: foo.x bar.x
               foo.x bar.x: newer.x
                       @echo Building $@
               % touch ../bar.x
               % touch ../newer.x
               % make        # GNU make
               Building foo.x
               Building bar.x
               % pmake       # NetBSD make
               Building foo.x
               Building ../bar.x
               % fmake       # FreeBSD make, OpenBSD make
               Building foo.x
               Building bar.x
               % tmake       # Tru64 make
               Building foo.x
               Building bar.x
               % touch ../bar.x
               % make        # GNU make
               Building foo.x
               % pmake       # NetBSD make
               Building foo.x
               % fmake       # FreeBSD make, OpenBSD make
               Building foo.x
               Building bar.x
               % tmake       # Tru64 make
               Building foo.x
               Building bar.x

Note how NetBSD make updates ../bar.x in its VPATH location, and how FreeBSD, OpenBSD, and Tru64 make always update bar.x, even when ../bar.x is up to date.

Another point worth mentioning is that once GNU make has decided to ignore a VPATH filename (e.g., it ignored ../bar.x in the above example) it will continue to ignore it when the target occurs as a prerequisite of another rule.

The following example shows that GNU make does not look up bar.x in VPATH before performing the .x.y rule, because it ignored the VPATH result of bar.x while running the bar.x: newer.x rule.

               % cat Makefile
               VPATH = ..
               all: bar.y
               bar.x: newer.x
                       @echo Building $@
               .SUFFIXES: .x .y
                       cp $< $@
               % touch ../bar.x
               % touch ../newer.x
               % make        # GNU make
               Building bar.x
               cp bar.x bar.y
               cp: cannot stat `bar.x': No such file or directory
               make: *** [bar.y] Error 1
               % pmake       # NetBSD make
               Building ../bar.x
               cp ../bar.x bar.y
               % rm bar.y
               % fmake       # FreeBSD make, OpenBSD make
               echo Building bar.x
               cp bar.x bar.y
               cp: cannot stat `bar.x': No such file or directory
               *** Error code 1
               % tmake       # Tru64 make
               Building bar.x
               cp: bar.x: No such file or directory
               *** Exit 1

Note that if you drop away the command from the bar.x: newer.x rule, GNU make will magically start to work: it knows that bar.x hasn't been updated, therefore it doesn't discard the result from VPATH (../bar.x) in succeeding uses. Tru64 will also work, but FreeBSD and OpenBSD still don't.

               % cat Makefile
               VPATH = ..
               all: bar.y
               bar.x: newer.x
               .SUFFIXES: .x .y
                       cp $< $@
               % touch ../bar.x
               % touch ../newer.x
               % make        # GNU make
               cp ../bar.x bar.y
               % rm bar.y
               % pmake       # NetBSD make
               cp ../bar.x bar.y
               % rm bar.y
               % fmake       # FreeBSD make, OpenBSD make
               cp bar.x bar.y
               cp: cannot stat `bar.x': No such file or directory
               *** Error code 1
               % tmake       # True64 make
               cp ../bar.x bar.y

It seems the sole solution that would please every make implementation is to never rely on VPATH searches for targets. In other words, VPATH should be reserved to unbuilt sources.

Single Suffix Rules and Separated Dependencies
A Single Suffix Rule is basically a usual suffix (inference) rule (, but which destination suffix is empty (.from:).

Separated dependencies simply refers to listing the prerequisite of a target, without defining a rule. Usually one can list on the one hand side, the rules, and on the other hand side, the dependencies.

Solaris make does not support separated dependencies for targets defined by single suffix rules:

          $ cat Makefile
          .SUFFIXES: .in
                  cp $< $ $ touch
          $ make
          $ ls

while GNU Make does:

          $ gmake
          cp foo
          $ ls
          Makefile  foo

Note it works without the foo: dependency.

          $ cat Makefile
          .SUFFIXES: .in
                  cp $< $ $ make foo
          cp foo

and it works with double suffix inference rules:

          $ cat Makefile
          .SUFFIXES: .in .out
                  cp $< $ $ make
          cp foo.out

As a result, in such a case, you have to write target rules.

Timestamp Resolution
Traditionally, file timestamps had 1-second resolution, and make used those timestamps to determine whether one file was newer than the other. However, many modern filesystems have timestamps with 1-nanosecond resolution. Some make implementations look at the entire timestamp; others ignore the fractional part, which can lead to incorrect results. Normally this is not a problem, but in some extreme cases you may need to use tricks like sleep 1 to work around timestamp truncation bugs.

Commands like cp -p and touch -r typically do not copy file timestamps to their full resolutions (see Limitations of Usual Tools). Hence you should be wary of rules like this:

          dest: src
                  cp -p src dest

as dest will often appear to be older than src after the timestamp is truncated, and this can cause make to do needless rework the next time it is invoked. To work around this problem, you can use a timestamp file, e.g.:

          dest-stamp: src
                  cp -p src dest
                  date >dest-stamp

11 Manual Configuration

A few kinds of features can't be guessed automatically by running test programs. For example, the details of the object-file format, or special options that need to be passed to the compiler or linker. You can check for such features using ad-hoc means, such as having configure check the output of the uname program, or looking for libraries that are unique to particular systems. However, Autoconf provides a uniform method for handling unguessable features.

11.1 Specifying the System Type

Like other GNU configure scripts, Autoconf-generated configure scripts can make decisions based on a canonical name for the system type, which has the form: cpu-vendor-os, where os can be system or kernel-system

configure can usually guess the canonical name for the type of system it's running on. To do so it runs a script called config.guess, which infers the name using the uname command or symbols predefined by the C preprocessor.

Alternately, the user can specify the system type with command line arguments to configure. Doing so is necessary when cross-compiling. In the most complex case of cross-compiling, three system types are involved. The options to specify them are:

the type of system on which the package is being configured and compiled. It defaults to the result of running config.guess.
the type of system on which the package will run. By default it is the same as the build machine. Specifying it enables the cross-compilation mode.
the type of system for which any compiler tools in the package will produce code (rarely needed). By default, it is the same as host.

If you mean to override the result of config.guess, use --build, not --host, since the latter enables cross-compilation. For historical reasons, passing --host also changes the build type. Therefore, whenever you specify --host, be sure to specify --build too. This will be fixed in the future.

     ./configure --build=i686-pc-linux-gnu --host=m68k-coff

will enter cross-compilation mode, but configure will fail if it can't run the code generated by the specified compiler if you configure as follows:

     ./configure CC=m68k-coff-gcc

configure recognizes short aliases for many system types; for example, decstation can be used instead of mips-dec-ultrix4.2. configure runs a script called config.sub to canonicalize system type aliases.

This section deliberately omits the description of the obsolete interface; see Hosts and Cross-Compilation.

11.2 Getting the Canonical System Type

The following macros make the system type available to configure scripts.

The variables build_alias, host_alias, and target_alias are always exactly the arguments of --build, --host, and --target; in particular, they are left empty if the user did not use them, even if the corresponding AC_CANONICAL macro was run. Any configure script may use these variables anywhere. These are the variables that should be used when in interaction with the user.

If you need to recognize some special environments based on their system type, run the following macros to get canonical system names. These variables are not set before the macro call.

If you use these macros, you must distribute config.guess and config.sub along with your source code. See Output, for information about the AC_CONFIG_AUX_DIR macro which you can use to control in which directory configure looks for those scripts.


Compute the canonical build-system type variable, build, and its three individual parts build_cpu, build_vendor, and build_os.

If --build was specified, then build is the canonicalization of build_alias by config.sub, otherwise it is determined by the shell script config.guess.


Compute the canonical host-system type variable, host, and its three individual parts host_cpu, host_vendor, and host_os.

If --host was specified, then host is the canonicalization of host_alias by config.sub, otherwise it defaults to build.


Compute the canonical target-system type variable, target, and its three individual parts target_cpu, target_vendor, and target_os.

If --target was specified, then target is the canonicalization of target_alias by config.sub, otherwise it defaults to host.

Note that there can be artifacts due to the backward compatibility code. See See Hosts and Cross-Compilation, for more.

11.3 Using the System Type

How do you use a canonical system type? Usually, you use it in one or more case statements in to select system-specific C files. Then, using AC_CONFIG_LINKS, link those files which have names based on the system name, to generic names, such as host.h or target.c (see Configuration Links). The case statement patterns can use shell wild cards to group several cases together, like in this fragment:

     case $target in
     i386-*-mach* | i386-*-gnu*)
                  obj_format=aout emulation=mach bfd_gas=yes ;;
     i960-*-bout) obj_format=bout ;;

and later in, use:


Note that the above example uses $target because it's taken from a tool which can be built on some architecture ($build), run on another ($host), but yet handle data for a third architecture ($target). Such tools are usually part of a compiler suite, they generate code for a specific $target.

However $target should be meaningless for most packages. If you want to base a decision on the system where your program will be run, make sure you use the $host variable, as in the following excerpt:

     case $host in
       *-*-msdos* | *-*-go32* | *-*-mingw32* | *-*-cygwin* | *-*-windows*)

You can also use the host system type to find cross-compilation tools. See Generic Programs, for information about the AC_CHECK_TOOL macro which does that.

12 Site Configuration

configure scripts support several kinds of local configuration decisions. There are ways for users to specify where external software packages are, include or exclude optional features, install programs under modified names, and set default values for configure options.

12.1 Working With External Software

Some packages require, or can optionally use, other software packages that are already installed. The user can give configure command line options to specify which such external software to use. The options have one of these forms:


For example, --with-gnu-ld means work with the GNU linker instead of some other linker. --with-x means work with The X Window System.

The user can give an argument by following the package name with = and the argument. Giving an argument of no is for packages that are used by default; it says to not use the package. An argument that is neither yes nor no could include a name or number of a version of the other package, to specify more precisely which other package this program is supposed to work with. If no argument is given, it defaults to yes. --without-package is equivalent to --with-package=no.

configure scripts do not complain about --with-package options that they do not support. This behavior permits configuring a source tree containing multiple packages with a top-level configure script when the packages support different options, without spurious error messages about options that some of the packages support. An unfortunate side effect is that option spelling errors are not diagnosed. No better approach to this problem has been suggested so far.

For each external software package that may be used, should call AC_ARG_WITH to detect whether the configure user asked to use it. Whether each package is used or not by default, and which arguments are valid, is up to you.

— Macro: AC_ARG_WITH (package, help-string, [action-if-given], [action-if-not-given])

If the user gave configure the option --with-package or --without-package, run shell commands action-if-given. If neither option was given, run shell commands action-if-not-given. The name package indicates another software package that this program should work with. It should consist only of alphanumeric characters and dashes.

The option's argument is available to the shell commands action-if-given in the shell variable withval, which is actually just the value of the shell variable with_package, with any - characters changed into _. You may use that variable instead, if you wish.

The argument help-string is a description of the option that looks like this:

            --with-readline         support fancy command line editing

help-string may be more than one line long, if more detail is needed. Just make sure the columns line up in configure --help. Avoid tabs in the help string. You'll need to enclose the help string in [ and ] in order to produce the leading spaces.

You should format your help-string with the macro AS_HELP_STRING (see Pretty Help Strings).

— Macro: AC_WITH (package, action-if-given, [action-if-not-given])

This is an obsolete version of AC_ARG_WITH that does not support providing a help string.

12.2 Choosing Package Options

If a software package has optional compile-time features, the user can give configure command line options to specify whether to compile them. The options have one of these forms:


These options allow users to choose which optional features to build and install. --enable-feature options should never make a feature behave differently or cause one feature to replace another. They should only cause parts of the program to be built rather than left out.

The user can give an argument by following the feature name with = and the argument. Giving an argument of no requests that the feature not be made available. A feature with an argument looks like --enable-debug=stabs. If no argument is given, it defaults to yes. --disable-feature is equivalent to --enable-feature=no.

configure scripts do not complain about --enable-feature options that they do not support. This behavior permits configuring a source tree containing multiple packages with a top-level configure script when the packages support different options, without spurious error messages about options that some of the packages support. An unfortunate side effect is that option spelling errors are not diagnosed. No better approach to this problem has been suggested so far.

For each optional feature, should call AC_ARG_ENABLE to detect whether the configure user asked to include it. Whether each feature is included or not by default, and which arguments are valid, is up to you.

— Macro: AC_ARG_ENABLE (feature, help-string, [action-if-given], [action-if-not-given])

If the user gave configure the option --enable-feature or --disable-feature, run shell commands action-if-given. If neither option was given, run shell commands action-if-not-given. The name feature indicates an optional user-level facility. It should consist only of alphanumeric characters and dashes.

The option's argument is available to the shell commands action-if-given in the shell variable enableval, which is actually just the value of the shell variable enable_feature, with any - characters changed into _. You may use that variable instead, if you wish. The help-string argument is like that of AC_ARG_WITH (see External Software).

You should format your help-string with the macro AS_HELP_STRING (see Pretty Help Strings).

— Macro: AC_ENABLE (feature, action-if-given, [action-if-not-given])

This is an obsolete version of AC_ARG_ENABLE that does not support providing a help string.

12.3 Making Your Help Strings Look Pretty

Properly formatting the help strings which are used in AC_ARG_WITH (see External Software) and AC_ARG_ENABLE (see Package Options) can be challenging. Specifically, you want your own help strings to line up in the appropriate columns of configure --help just like the standard Autoconf help strings do. This is the purpose of the AS_HELP_STRING macro.

— Macro: AS_HELP_STRING (left-hand-side, right-hand-side)

Expands into an help string that looks pretty when the user executes configure --help. It is typically used in AC_ARG_WITH (see External Software) or AC_ARG_ENABLE (see Package Options). The following example will make this clearer.

                                      [use foo (default is NO)]),
                       [ac_cv_use_foo=$withval], [ac_cv_use_foo=no])
          AC_CACHE_CHECK([whether to use foo],
                         [ac_cv_use_foo], [ac_cv_use_foo=no])])

Please note that the call to AS_HELP_STRING is unquoted. Then the last few lines of configure --help will appear like this:

          --enable and --with options recognized:
            --with-foo              use foo (default is NO)

The AS_HELP_STRING macro is particularly helpful when the left-hand-side and/or right-hand-side are composed of macro arguments, as shown in the following example.

                       AS_HELP_STRING([--with-$1], [use $1 (default is $2)]),
                       ac_cv_use_$1=$withval, ac_cv_use_$1=no),
          AC_CACHE_CHECK(whether to use $1, ac_cv_use_$1, ac_cv_use_$1=$2)])

12.4 Configuring Site Details

Some software packages require complex site-specific information. Some examples are host names to use for certain services, company names, and email addresses to contact. Since some configuration scripts generated by Metaconfig ask for such information interactively, people sometimes wonder how to get that information in Autoconf-generated configuration scripts, which aren't interactive.

Such site configuration information should be put in a file that is edited only by users, not by programs. The location of the file can either be based on the prefix variable, or be a standard location such as the user's home directory. It could even be specified by an environment variable. The programs should examine that file at run time, rather than at compile time. Run-time configuration is more convenient for users and makes the configuration process simpler than getting the information while configuring. See Variables for Installation Directories (GNU Coding Standards), for more information on where to put data files.

12.5 Transforming Program Names When Installing

Autoconf supports changing the names of programs when installing them. In order to use these transformations, must call the macro AC_ARG_PROGRAM.


Place in output variable program_transform_name a sequence of sed commands for changing the names of installed programs.

If any of the options described below are given to configure, program names are transformed accordingly. Otherwise, if AC_CANONICAL_TARGET has been called and a --target value is given, the target type followed by a dash is used as a prefix. Otherwise, no program name transformation is done.

12.5.1 Transformation Options

You can specify name transformations by giving configure these command line options:

prepend prefix to the names;
append suffix to the names;
perform sed substitution expression on the names.

12.5.2 Transformation Examples

These transformations are useful with programs that can be part of a cross-compilation development environment. For example, a cross-assembler running on a Sun 4 configured with --target=i960-vxworks is normally installed as i960-vxworks-as, rather than as, which could be confused with a native Sun 4 assembler.

You can force a program name to begin with g, if you don't want GNU programs installed on your system to shadow other programs with the same name. For example, if you configure GNU diff with --program-prefix=g, then when you run make install it is installed as /usr/local/bin/gdiff.

As a more sophisticated example, you could use

     --program-transform-name='s/^/g/; s/^gg/g/; s/^gless/less/'

to prepend g to most of the program names in a source tree, excepting those like gdb that already have one and those like less and lesskey that aren't GNU programs. (That is assuming that you have a source tree containing those programs that is set up to use this feature.)

One way to install multiple versions of some programs simultaneously is to append a version number to the name of one or both. For example, if you want to keep Autoconf version 1 around for awhile, you can configure Autoconf version 2 using --program-suffix=2 to install the programs as /usr/local/bin/autoconf2, /usr/local/bin/autoheader2, etc. Nevertheless, pay attention that only the binaries are renamed, therefore you'd have problems with the library files which might overlap.

12.5.3 Transformation Rules

Here is how to use the variable program_transform_name in a

     PROGRAMS = cp ls rm
     transform = @program_transform_name@
             for p in $(PROGRAMS); do \
               $(INSTALL_PROGRAM) $$p $(DESTDIR)$(bindir)/`echo $$p | \
                                                   sed '$(transform)'`; \
             for p in $(PROGRAMS); do \
               rm -f $(DESTDIR)$(bindir)/`echo $$p | sed '$(transform)'`; \

It is guaranteed that program_transform_name is never empty, and that there are no useless separators. Therefore you may safely embed program_transform_name within a sed program using ;:

     transform = @program_transform_name@
     transform_exe = s/$(EXEEXT)$$//;$(transform);s/$$/$(EXEEXT)/

Whether to do the transformations on documentation files (Texinfo or man) is a tricky question; there seems to be no perfect answer, due to the several reasons for name transforming. Documentation is not usually particular to a specific architecture, and Texinfo files do not conflict with system documentation. But they might conflict with earlier versions of the same files, and man pages sometimes do conflict with system documentation. As a compromise, it is probably best to do name transformations on man pages but not on Texinfo manuals.

12.6 Setting Site Defaults

Autoconf-generated configure scripts allow your site to provide default values for some configuration values. You do this by creating site- and system-wide initialization files.

If the environment variable CONFIG_SITE is set, configure uses its value as the name of a shell script to read. Otherwise, it reads the shell script prefix/share/ if it exists, then prefix/etc/ if it exists. Thus, settings in machine-specific files override those in machine-independent ones in case of conflict.

Site files can be arbitrary shell scripts, but only certain kinds of code are really appropriate to be in them. Because configure reads any cache file after it has read any site files, a site file can define a default cache file to be shared between all Autoconf-generated configure scripts run on that system (see Cache Files). If you set a default cache file in a site file, it is a good idea to also set the output variable CC in that site file, because the cache file is only valid for a particular compiler, but many systems have several available.

You can examine or override the value set by a command line option to configure in a site file; options set shell variables that have the same names as the options, with any dashes turned into underscores. The exceptions are that --without- and --disable- options are like giving the corresponding --with- or --enable- option and the value no. Thus, --cache-file=localcache sets the variable cache_file to the value localcache; --enable-warnings=no or --disable-warnings sets the variable enable_warnings to the value no; --prefix=/usr sets the variable prefix to the value /usr; etc.

Site files are also good places to set default values for other output variables, such as CFLAGS, if you need to give them non-default values: anything you would normally do, repetitively, on the command line. If you use non-default values for prefix or exec_prefix (wherever you locate the site file), you can set them in the site file if you specify it with the CONFIG_SITE environment variable.

You can set some cache values in the site file itself. Doing this is useful if you are cross-compiling, where it is impossible to check features that require running a test program. You could “prime the cache” by setting those values correctly for that system in prefix/etc/ To find out the names of the cache variables you need to set, look for shell variables with _cv_ in their names in the affected configure scripts, or in the Autoconf M4 source code for those macros.

The cache file is careful to not override any variables set in the site files. Similarly, you should not override command-line options in the site files. Your code should check that variables such as prefix and cache_file have their default values (as set near the top of configure) before changing them.

Here is a sample file /usr/share/local/gnu/share/ The command configure --prefix=/usr/share/local/gnu would read this file (if CONFIG_SITE is not set to a different file).

     # for configure
     # Change some defaults.
     test "$prefix" = NONE && prefix=/usr/share/local/gnu
     test "$exec_prefix" = NONE && exec_prefix=/usr/local/gnu
     test "$sharedstatedir" = '$prefix/com' && sharedstatedir=/var
     test "$localstatedir" = '$prefix/var' && localstatedir=/var
     # Give Autoconf 2.x generated configure scripts a shared default
     # cache file for feature test results, architecture-specific.
     if test "$cache_file" = /dev/null; then
       # A cache file is only valid for one C compiler.

13 Running configure Scripts

Below are instructions on how to configure a package that uses a configure script, suitable for inclusion as an INSTALL file in the package. A plain-text version of INSTALL which you may use comes with Autoconf.

13.1 Basic Installation

These are generic installation instructions.

The configure shell script attempts to guess correct values for various system-dependent variables used during compilation. It uses those values to create a Makefile in each directory of the package. It may also create one or more .h files containing system-dependent definitions. Finally, it creates a shell script config.status that you can run in the future to recreate the current configuration, and a file config.log containing compiler output (useful mainly for debugging configure).

It can also use an optional file (typically called config.cache and enabled with --cache-file=config.cache or simply -C) that saves the results of its tests to speed up reconfiguring. (Caching is disabled by default to prevent problems with accidental use of stale cache files.)

If you need to do unusual things to compile the package, please try to figure out how configure could check whether to do them, and mail diffs or instructions to the address given in the README so they can be considered for the next release. If you are using the cache, and at some point config.cache contains results you don't want to keep, you may remove or edit it.

The file (or is used to create configure by a program called autoconf. You only need if you want to change it or regenerate configure using a newer version of autoconf.

The simplest way to compile this package is:

  1. cd to the directory containing the package's source code and type ./configure to configure the package for your system. If you're using csh on an old version of System V, you might need to type sh ./configure instead to prevent csh from trying to execute configure itself.

    Running configure takes awhile. While running, it prints some messages telling which features it is checking for.

  2. Type make to compile the package.
  3. Optionally, type make check to run any self-tests that come with the package.
  4. Type make install to install the programs and any data files and documentation.
  5. You can remove the program binaries and object files from the source code directory by typing make clean. To also remove the files that configure created (so you can compile the package for a different kind of computer), type make distclean. There is also a make maintainer-clean target, but that is intended mainly for the package's developers. If you use it, you may have to get all sorts of other programs in order to regenerate files that came with the distribution.

13.2 Compilers and Options

Some systems require unusual options for compilation or linking that the configure script does not know about. Run ./configure --help for details on some of the pertinent environment variables.

You can give configure initial values for configuration parameters by setting variables in the command line or in the environment. Here is an example:

     ./configure CC=c89 CFLAGS=-O2 LIBS=-lposix

See Defining Variables, for more details.

13.3 Compiling For Multiple Architectures

You can compile the package for more than one kind of computer at the same time, by placing the object files for each architecture in their own directory. To do this, you must use a version of make that supports the VPATH variable, such as GNU make. cd to the directory where you want the object files and executables to go and run the configure script. configure automatically checks for the source code in the directory that configure is in and in ...

If you have to use a make that does not support the VPATH variable, you have to compile the package for one architecture at a time in the source code directory. After you have installed the package for one architecture, use make distclean before reconfiguring for another architecture.

13.4 Installation Names

By default, make install will install the package's files in /usr/local/bin, /usr/local/man, etc. You can specify an installation prefix other than /usr/local by giving configure the option --prefix=path.

You can specify separate installation prefixes for architecture-specific files and architecture-independent files. If you give configure the option --exec-prefix=path, the package will use path as the prefix for installing programs and libraries. Documentation and other data files will still use the regular prefix.

In addition, if you use an unusual directory layout you can give options like --bindir=path to specify different values for particular kinds of files. Run configure --help for a list of the directories you can set and what kinds of files go in them.

If the package supports it, you can cause programs to be installed with an extra prefix or suffix on their names by giving configure the option --program-prefix=PREFIX or --program-suffix=SUFFIX.

13.5 Optional Features

Some packages pay attention to --enable-feature options to configure, where feature indicates an optional part of the package. They may also pay attention to --with-package options, where package is something like gnu-as or x (for the X Window System). The README should mention any --enable- and --with- options that the package recognizes.

For packages that use the X Window System, configure can usually find the X include and library files automatically, but if it doesn't, you can use the configure options --x-includes=dir and --x-libraries=dir to specify their locations.

13.6 Specifying the System Type

There may be some features configure cannot figure out automatically, but needs to determine by the type of machine the package will run on. Usually, assuming the package is built to be run on the same architectures, configure can figure that out, but if it prints a message saying it cannot guess the machine type, give it the --build=type option. type can either be a short name for the system type, such as sun4, or a canonical name which has the form:


where system can have one of these forms:

     os kernel-os

See the file config.sub for the possible values of each field. If config.sub isn't included in this package, then this package doesn't need to know the machine type.

If you are building compiler tools for cross-compiling, you should use the --target=type option to select the type of system they will produce code for.

If you want to use a cross compiler, that generates code for a platform different from the build platform, you should specify the host platform (i.e., that on which the generated programs will eventually be run) with --host=type.

13.7 Sharing Defaults

If you want to set default values for configure scripts to share, you can create a site shell script called that gives default values for variables like CC, cache_file, and prefix. configure looks for prefix/share/ if it exists, then prefix/etc/ if it exists. Or, you can set the CONFIG_SITE environment variable to the location of the site script. A warning: not all configure scripts look for a site script.

13.8 Defining Variables

Variables not defined in a site shell script can be set in the environment passed to configure. However, some packages may run configure again during the build, and the customized values of these variables may be lost. In order to avoid this problem, you should set them in the configure command line, using VAR=value. For example:

     ./configure CC=/usr/local2/bin/gcc

will cause the specified gcc to be used as the C compiler (unless it is overridden in the site shell script).

13.9 configure Invocation

configure recognizes the following options to control how it operates.

Print a summary of the options to configure, and exit.
Print the version of Autoconf used to generate the configure script, and exit.
Enable the cache: use and save the results of the tests in file, traditionally config.cache. file defaults to /dev/null to disable caching.
Alias for --cache-file=config.cache.
Do not print messages saying which checks are being made. To suppress all normal output, redirect it to /dev/null (any error messages will still be shown).
Look for the package's source code in directory dir. Usually configure can determine that directory automatically.

configure also accepts some other, not widely useful, options. Run configure --help for more details.

14 Recreating a Configuration

The configure script creates a file named config.status, which actually configures, instantiates, the template files. It also records the configuration options that were specified when the package was last configured in case reconfiguring is needed.


     ./config.status option... [file...]

It configures the files; if none are specified, all the templates are instantiated. The files must be specified without their dependencies, as in

     ./config.status foobar



The supported options are:

Print a summary of the command line options, the list of the template files, and exit.
Print the version number of Autoconf and exit.
Do not print progress messages.
Don't remove the temporary files.
Require that file be instantiated as if AC_CONFIG_FILES(file:template) was used. Both file and template may be - in which case the standard output and/or standard input, respectively, is used. If a template filename is relative, it is first looked for in the build tree, and then in the source tree. See Configuration Actions, for more details.

This option and the following ones provide one way for separately distributed packages to share the values computed by configure. Doing so can be useful if some of the packages need a superset of the features that one of them, perhaps a common library, does. These options allow a config.status file to create files other than the ones that its specifies, so it can be used for a different package.

Same as --file above, but with AC_CONFIG_HEADERS.
Ask config.status to update itself and exit (no instantiation). This option is useful if you change configure, so that the results of some tests might be different from the previous run. The --recheck option re-runs configure with the same arguments you used before, plus the --no-create option, which prevents configure from running config.status and creating Makefile and other files, and the --no-recursion option, which prevents configure from running other configure scripts in subdirectories. (This is so other Makefile rules can run config.status when it changes; see Automatic Remaking, for an example).

config.status checks several optional environment variables that can alter its behavior:

— Variable: CONFIG_SHELL

The shell with which to run configure for the --recheck option. It must be Bourne-compatible. The default is a shell that supports LINENO if available, and /bin/sh otherwise.


The file name to use for the shell script that records the configuration. The default is ./config.status. This variable is useful when one package uses parts of another and the configure scripts shouldn't be merged because they are maintained separately.

You can use ./config.status in your Makefiles. For example, in the dependencies given above (see Automatic Remaking), config.status is run twice when has changed. If that bothers you, you can make each run only regenerate the files for that rule:

     config.h: stamp-h
     stamp-h: config.status
             ./config.status config.h
             echo > stamp-h
     Makefile: config.status
             ./config.status Makefile

The calling convention of config.status has changed; see Obsolete config.status Use, for details.

15 Obsolete Constructs

Autoconf changes, and throughout the years some constructs have been obsoleted. Most of the changes involve the macros, but in some cases the tools themselves, or even some concepts, are now considered obsolete.

You may completely skip this chapter if you are new to Autoconf. Its intention is mainly to help maintainers updating their packages by understanding how to move to more modern constructs.

15.1 Obsolete config.status Invocation

config.status now supports arguments to specify the files to instantiate; see config.status Invocation, for more details. Before, environment variables had to be used.


The tags of the commands to execute. The default is the arguments given to AC_OUTPUT and AC_CONFIG_COMMANDS in

— Variable: CONFIG_FILES

The files in which to perform @variable@ substitutions. The default is the arguments given to AC_OUTPUT and AC_CONFIG_FILES in


The files in which to substitute C #define statements. The default is the arguments given to AC_CONFIG_HEADERS; if that macro was not called, config.status ignores this variable.

— Variable: CONFIG_LINKS

The symbolic links to establish. The default is the arguments given to AC_CONFIG_LINKS; if that macro was not called, config.status ignores this variable.

In config.status Invocation, using this old interface, the example would be:

     config.h: stamp-h
     stamp-h: config.status
               CONFIG_HEADERS=config.h ./config.status
             echo > stamp-h
     Makefile: config.status
               CONFIG_FILES=Makefile ./config.status

(If does not call AC_CONFIG_HEADERS, there is no need to set CONFIG_HEADERS in the make rules. Equally for CONFIG_COMMANDS etc.)

15.2 acconfig.h

In order to produce, autoheader needs to build or to find templates for each symbol. Modern releases of Autoconf use AH_VERBATIM and AH_TEMPLATE (see Autoheader Macros), but in older releases a file, acconfig.h, contained the list of needed templates. autoheader copied comments and #define and #undef statements from acconfig.h in the current directory, if present. This file used to be mandatory if you AC_DEFINE any additional symbols.

Modern releases of Autoconf also provide AH_TOP and AH_BOTTOM if you need to prepend/append some information to Ancient versions of Autoconf had a similar feature: if ./acconfig.h contains the string @TOP@, autoheader copies the lines before the line containing @TOP@ into the top of the file that it generates. Similarly, if ./acconfig.h contains the string @BOTTOM@, autoheader copies the lines after that line to the end of the file it generates. Either or both of those strings may be omitted. An even older alternate way to produce the same effect in ancient versions of Autoconf is to create the files (typically and/or in the current directory. If they exist, autoheader copies them to the beginning and end, respectively, of its output.

In former versions of Autoconf, the files used in preparing a software package for distribution were: --.   .------> autoconf* -----> configure
     [aclocal.m4] --+   `---.
     [acsite.m4] ---'       |
                            +--> [autoheader*] -> []
     [acconfig.h] ----.     |
     [] --+
     [] --'

Using only the AH_ macros, should be self-contained, and should not depend upon acconfig.h etc.

15.3 Using autoupdate to Modernize

The autoupdate program updates a file that calls Autoconf macros by their old names to use the current macro names. In version 2 of Autoconf, most of the macros were renamed to use a more uniform and descriptive naming scheme. See Macro Names, for a description of the new scheme. Although the old names still work (see Obsolete Macros, for a list of the old macros and the corresponding new names), you can make your files more readable and make it easier to use the current Autoconf documentation if you update them to use the new macro names.

If given no arguments, autoupdate updates, backing up the original version with the suffix ~ (or the value of the environment variable SIMPLE_BACKUP_SUFFIX, if that is set). If you give autoupdate an argument, it updates that file instead of

autoupdate accepts the following options:

Print a summary of the command line options and exit.
Print the version number of Autoconf and exit.
Report processing steps.
Don't remove the temporary files.
Force the update even if the file has not changed. Disregard the cache.
-I dir
Also look for input files in dir. Multiple invocations accumulate. Directories are browsed from last to first.

15.4 Obsolete Macros

Several macros are obsoleted in Autoconf, for various reasons (typically they failed to quote properly, couldn't be extended for more recent issues etc.). They are still supported, but deprecated: their use should be avoided.

During the jump from Autoconf version 1 to version 2, most of the macros were renamed to use a more uniform and descriptive naming scheme, but their signature did not change. See Macro Names, for a description of the new naming scheme. Below, if there is just the mapping from old names to new names for these macros, the reader is invited to refer to the definition of the new macro for the signature and the description.

— Macro: AC_ALLOCA



removed because of limited usefulness

— Macro: AC_C_CROSS

This macro is obsolete; it does nothing.


Determine the system type and set output variables to the names of the canonical system types. See Canonicalizing, for details about the variables this macro sets.

The user is encouraged to use either AC_CANONICAL_BUILD, or AC_CANONICAL_HOST, or AC_CANONICAL_TARGET, depending on the needs. Using AC_CANONICAL_TARGET is enough to run the two other macros.



— Macro: AC_CHECK_TYPE (type, default)

Autoconf, up to 2.13, used to provide this version of AC_CHECK_TYPE, deprecated because of its flaws. Firstly, although it is a member of the CHECK clan, singular sub-family, it does more than just checking. Secondly, missing types are not typedef'd, they are #define'd, which can lead to incompatible code in the case of pointer types.

This use of AC_CHECK_TYPE is obsolete and discouraged; see Generic Types, for the description of the current macro.

If the type type is not defined, define it to be the C (or C++) builtin type default, e.g., short or unsigned.

This macro is equivalent to:

                        [AC_DEFINE_UNQUOTED([type], [default],
                                            [Define to `default' if
                                             <sys/types.h> does not define.])])

In order to keep backward compatibility, the two versions of AC_CHECK_TYPE are implemented, selected by a simple heuristics:

  1. If there are three or four arguments, the modern version is used.
  2. If the second argument appears to be a C or C++ type, then the obsolete version is used. This happens if the argument is a C or C++ builtin type or a C identifier ending in _t, optionally followed by one of [(* and then by a string of zero or more characters taken from the set []()* _a-zA-Z0-9.
  3. If the second argument is spelled with the alphabet of valid C and C++ types, the user is warned and the modern version is used.
  4. Otherwise, the modern version is used.

You are encouraged either to use a valid builtin type, or to use the equivalent modern code (see above), or better yet, to use AC_CHECK_TYPES together with

          #if !HAVE_LOFF_T
          typedef loff_t off_t;
— Macro: AC_CHECKING (feature-description)

Same as AC_MSG_NOTICE([checking feature-description...].

— Macro: AC_COMPILE_CHECK (echo-text, includes, function-body, action-if-found, [action-if-not-found])

This is an obsolete version of AC_TRY_COMPILE itself replaced by AC_COMPILE_IFELSE (see Running the Compiler), with the addition that it prints checking for echo-text to the standard output first, if echo-text is non-empty. Use AC_MSG_CHECKING and AC_MSG_RESULT instead to print messages (see Printing Messages).

— Macro: AC_CONST



Same as AC_C_CROSS, which is obsolete too, and does nothing :-).

— Macro: AC_CYGWIN

Check for the Cygwin environment in which case the shell variable CYGWIN is set to yes. Don't use this macro, the dignified means to check the nature of the host is using AC_CANONICAL_HOST. As a matter of fact this macro is defined as:

          case $host_os in
            *cygwin* ) CYGWIN=yes;;
                   * ) CYGWIN=no;;

Beware that the variable CYGWIN has a very special meaning when running CygWin32, and should not be changed. That's yet another reason not to use this macro.


Same as:

          [#include <signal.h>
          /* NetBSD declares sys_siglist in unistd.h.  */
          #if HAVE_UNISTD_H
          # include <unistd.h>

Does nothing, now integrated in AC_PROG_LEX.


Like calling AC_FUNC_CLOSEDIR_VOID andAC_HEADER_DIRENT, but defines a different set of C preprocessor macros to indicate which header file is found:

Header Old Symbol New Symbol


If on DYNIX/ptx, add -lseq to output variable LIBS. This macro used to be defined as

          AC_CHECK_LIB(seq, getmntent, LIBS="-lseq $LIBS")

now it is just AC_FUNC_GETMNTENT.

— Macro: AC_EXEEXT

Defined the output variable EXEEXT based on the output of the compiler, which is now done automatically. Typically set to empty string if Unix and .exe if Win32 or OS/2.

— Macro: AC_EMXOS2

Similar to AC_CYGWIN but checks for the EMX environment on OS/2 and sets EMXOS2.

— Macro: AC_ERROR


— Macro: AC_FIND_X






— Macro: AC_FUNC_WAIT3

If wait3 is found and fills in the contents of its third argument (a struct rusage *), which HP-UX does not do, define HAVE_WAIT3.

These days portable programs should use waitpid, not wait3, as wait3 is being removed from the Open Group standards, and will not appear in the next revision of POSIX.











— Macro: AC_HAVE_LIBRARY (library, [action-if-found], [action-if-not-found], [other-libraries])

This macro is equivalent to calling AC_CHECK_LIB with a function argument of main. In addition, library can be written as any of foo, -lfoo, or libfoo.a. In all of those cases, the compiler is passed -lfoo. However, library cannot be a shell variable; it must be a literal name.


AC_SYS_INTERPRETER (different calling convention)







— Macro: AC_INIT (unique-file-in-source-dir)

Formerly AC_INIT used to have a single argument, and was equivalent to:

— Macro: AC_INLINE


— Macro: AC_INT_16_BITS

If the C type int is 16 bits wide, define INT_16_BITS. Use AC_CHECK_SIZEOF(int) instead.

— Macro: AC_IRIX_SUN

If on irix (Silicon Graphics unix), add -lsun to output LIBS. If you were using it to get getmntent, use AC_FUNC_GETMNTENT instead. If you used it for the NIS versions of the password and group functions, use AC_CHECK_LIB(sun, getpwnam). Up to Autoconf 2.13, it used to be

          AC_CHECK_LIB(sun, getmntent, LIBS="-lsun $LIBS")

now it is defined as

          AC_CHECK_LIB(sun, getpwnam)
— Macro: AC_LANG_C

Same as AC_LANG(C).


Same as AC_LANG(C++).


Same as AC_LANG(Fortran 77).


Select the language that is saved on the top of the stack, as set by AC_LANG_SAVE, remove it from the stack, and call AC_LANG(language).


Remember the current language (as set by AC_LANG) on a stack. The current language does not change. AC_LANG_PUSH is preferred.

— Macro: AC_LINK_FILES (source..., dest...)

This is an obsolete version of AC_CONFIG_LINKS. An updated version of:

          AC_LINK_FILES(config/$machine.h config/$obj_format.h,
                        host.h            object.h)


— Macro: AC_LN_S


— Macro: AC_LONG_64_BITS

Define LONG_64_BITS if the C type long int is 64 bits wide. Use the generic macro AC_CHECK_SIZEOF([long int]) instead.







— Macro: AC_MEMORY_H

Used to define NEED_MEMORY_H if the mem functions were defined in memory.h. Today it is equivalent to AC_CHECK_HEADERS(memory.h). Adjust your code to depend upon HAVE_MEMORY_H, not NEED_MEMORY_H; see Standard Symbols.

— Macro: AC_MINGW32

Similar to AC_CYGWIN but checks for the MingW32 compiler environment and sets MINGW32.



— Macro: AC_MMAP


— Macro: AC_MODE_T


— Macro: AC_OBJEXT

Defined the output variable OBJEXT based on the output of the compiler, after .c files have been excluded. Typically set to o if Unix, obj if Win32. Now the compiler checking macros handle this automatically.

— Macro: AC_OBSOLETE (this-macro-name, [suggestion])

Make M4 print a message to the standard error output warning that this-macro-name is obsolete, and giving the file and line number where it was called. this-macro-name should be the name of the macro that is calling AC_OBSOLETE. If suggestion is given, it is printed at the end of the warning message; for example, it can be a suggestion for what to use instead of this-macro-name.

For instance

          AC_OBSOLETE([$0], [; use AC_CHECK_HEADERS(unistd.h) instead])dnl

You are encouraged to use AU_DEFUN instead, since it gives better services to the user.

— Macro: AC_OFF_T


— Macro: AC_OUTPUT ([file]..., [extra-cmds], [init-cmds])

The use of AC_OUTPUT with argument is deprecated. This obsoleted interface is equivalent to:

                             extra-cmds, init-cmds)
— Macro: AC_OUTPUT_COMMANDS (extra-cmds, [init-cmds])

Specify additional shell commands to run at the end of config.status, and shell commands to initialize any variables from configure. This macro may be called multiple times. It is obsolete, replaced by AC_CONFIG_COMMANDS.

Here is an unrealistic example:

          AC_OUTPUT_COMMANDS([echo this is extra $fubar, and so on.],
          AC_OUTPUT_COMMANDS([echo this is another, extra, bit],
                             [echo init bit])

Aside from the fact that AC_CONFIG_COMMANDS requires an additional key, an important difference is that AC_OUTPUT_COMMANDS is quoting its arguments twice, unlike AC_CONFIG_COMMANDS. This means that AC_CONFIG_COMMANDS can safely be given macro calls as arguments:

          AC_CONFIG_COMMANDS(foo, [my_FOO()])

Conversely, where one level of quoting was enough for literal strings with AC_OUTPUT_COMMANDS, you need two with AC_CONFIG_COMMANDS. The following lines are equivalent:

          AC_OUTPUT_COMMANDS([echo "Square brackets: []"])
          AC_CONFIG_COMMANDS([default], [[echo "Square brackets: []"]])
— Macro: AC_PID_T


— Macro: AC_PREFIX



This macro has been integrated into AC_PROG_CC.












removed because of limited usefulness





— Macro: AC_RSH

removed because of limited usefulness

— Macro: AC_SCO_INTL

If on SCO UNIX, add -lintl to output variable LIBS. This macro used to

          AC_CHECK_LIB(intl, strftime, LIBS="-lintl $LIBS")

Now it just calls AC_FUNC_STRFTIME instead.



— Macro: AC_SET_MAKE




— Macro: AC_SIZE_T












— Macro: AC_ST_RDEV



If the system automatically restarts a system call that is interrupted by a signal, define HAVE_RESTARTABLE_SYSCALLS. This macro does not check if system calls are restarted in general–it tests whether a signal handler installed with signal (but not sigaction) causes system calls to be restarted. It does not test if system calls can be restarted when interrupted by signals that have no handler.

These days portable programs should use sigaction with SA_RESTART if they want restartable system calls. They should not rely on HAVE_RESTARTABLE_SYSCALLS, since nowadays whether a system call is restartable is a dynamic issue, not a configuration-time issue.



— Macro: AC_TEST_CPP



AC_TRY_RUN, replaced by AC_RUN_IFELSE.





— Macro: AC_TRY_COMPILE (includes, function-body, [action-if-found], [action-if-not-found])

Same as AC_COMPILE_IFELSE([AC_LANG_SOURCE([[includes]], [[function-body]])], [action-if-true], [action-if-false]) (see Running the Compiler).

This macro double quotes both includes and function-body.

For C and C++, includes is any #include statements needed by the code in function-body (includes will be ignored if the currently selected language is Fortran or Fortran 77). The compiler and compilation flags are determined by the current language (see Language Choice).

— Macro: AC_TRY_CPP (input, [action-if-true], [action-if-false])

Same as AC_PREPROC_IFELSE([AC_LANG_SOURCE([[input]])], [action-if-true], [action-if-false]) (see Running the Preprocessor).

This macro double quotes the input.

— Macro: AC_TRY_LINK (includes, function-body, [action-if-found], [action-if-not-found])

Same as AC_LINK_IFELSE([AC_LANG_SOURCE([[includes]], [[function-body]])], [action-if-true], [action-if-false]) (see Running the Compiler).

This macro double quotes both includes and function-body.

Depending on the current language (see Language Choice), create a test program to see whether a function whose body consists of function-body can be compiled and linked. If the file compiles and links successfully, run shell commands action-if-found, otherwise run action-if-not-found.

This macro double quotes both includes and function-body.

For C and C++, includes is any #include statements needed by the code in function-body (includes will be ignored if the currently selected language is Fortran or Fortran 77). The compiler and compilation flags are determined by the current language (see Language Choice), and in addition LDFLAGS and LIBS are used for linking.

— Macro: AC_TRY_LINK_FUNC (function, [action-if-found], [action-if-not-found])

This macro is equivalent to AC_LINK_IFELSE([AC_LANG_CALL([[includes]], [[function-body]])], [action-if-true], [action-if-false]).

— Macro: AC_TRY_RUN (program, [action-if-true], [action-if-false], [action-if-cross-compiling])

Same as AC_RUN_IFELSE([AC_LANG_SOURCE([[program]], [action-if-true], [action-if-false], [action-if-cross-compiling]) (see Run Time).

— Macro: AC_UID_T


— Macro: AC_UNISTD_H

Same as AC_CHECK_HEADERS(unistd.h).

— Macro: AC_USG

Define USG if the BSD string functions are defined in strings.h. You should no longer depend upon USG, but on HAVE_STRING_H; see Standard Symbols.




If the cache file is inconsistent with the current host, target and build system types, it used to execute cmd or print a default error message. This is now handled by default.

— Macro: AC_VERBOSE (result-description)


— Macro: AC_VFORK




— Macro: AC_WAIT3


— Macro: AC_WARN





This macro used to add -lx to output variable LIBS if on Xenix. Also, if dirent.h is being checked for, added -ldir to LIBS. Now it is merely an alias of AC_HEADER_DIRENT instead, plus some code to detect whether running xenix on which you should not depend:

          AC_MSG_CHECKING([for Xenix])
          [#if defined M_XENIX && !defined M_UNIX
                       [AC_MSG_RESULT([yes]); XENIX=yes],
                       [AC_MSG_RESULT([no]); XENIX=])


15.5 Upgrading From Version 1

Autoconf version 2 is mostly backward compatible with version 1. However, it introduces better ways to do some things, and doesn't support some of the ugly things in version 1. So, depending on how sophisticated your files are, you might have to do some manual work in order to upgrade to version 2. This chapter points out some problems to watch for when upgrading. Also, perhaps your configure scripts could benefit from some of the new features in version 2; the changes are summarized in the file NEWS in the Autoconf distribution.

15.5.1 Changed File Names

If you have an aclocal.m4 installed with Autoconf (as opposed to in a particular package's source directory), you must rename it to acsite.m4. See autoconf Invocation.

If you distribute with your package, rename it to install-sh so make builtin rules won't inadvertently create a file called install from it. AC_PROG_INSTALL looks for the script under both names, but it is best to use the new name.

If you were using,, or acconfig.h, you still can, but you will have less clutter if you use the AH_ macros. See Autoheader Macros.

15.5.2 Changed Makefiles

Add @CFLAGS@, @CPPFLAGS@, and @LDFLAGS@ in your files, so they can take advantage of the values of those variables in the environment when configure is run. Doing this isn't necessary, but it's a convenience for users.

Also add @configure_input@ in a comment to each input file for AC_OUTPUT, so that the output files will contain a comment saying they were produced by configure. Automatically selecting the right comment syntax for all the kinds of files that people call AC_OUTPUT on became too much work.

Add config.log and config.cache to the list of files you remove in distclean targets.

If you have the following in

     prefix = /usr/local
     exec_prefix = $(prefix)

you must change it to:

     prefix = @prefix@
     exec_prefix = @exec_prefix@

The old behavior of replacing those variables without @ characters around them has been removed.

15.5.3 Changed Macros

Many of the macros were renamed in Autoconf version 2. You can still use the old names, but the new ones are clearer, and it's easier to find the documentation for them. See Obsolete Macros, for a table showing the new names for the old macros. Use the autoupdate program to convert your to using the new macro names. See autoupdate Invocation.

Some macros have been superseded by similar ones that do the job better, but are not call-compatible. If you get warnings about calling obsolete macros while running autoconf, you may safely ignore them, but your configure script will generally work better if you follow the advice that is printed about what to replace the obsolete macros with. In particular, the mechanism for reporting the results of tests has changed. If you were using echo or AC_VERBOSE (perhaps via AC_COMPILE_CHECK), your configure script's output will look better if you switch to AC_MSG_CHECKING and AC_MSG_RESULT. See Printing Messages. Those macros work best in conjunction with cache variables. See Caching Results.

15.5.4 Changed Results

If you were checking the results of previous tests by examining the shell variable DEFS, you need to switch to checking the values of the cache variables for those tests. DEFS no longer exists while configure is running; it is only created when generating output files. This difference from version 1 is because properly quoting the contents of that variable turned out to be too cumbersome and inefficient to do every time AC_DEFINE is called. See Cache Variable Names.

For example, here is a fragment written for Autoconf version 1:

     case "$DEFS" in
     *-DHAVE_SYSLOG*) ;;
     *) # syslog is not in the default libraries.  See if it's in some other.
       for lib in bsd socket inet; do
         AC_CHECKING(for syslog in -l$lib)
         LIBS="$saved_LIBS -l$lib"
         case "$DEFS" in
         *-DHAVE_SYSLOG*) break ;;
         *) ;;
       done ;;

Here is a way to write it for version 2:

     if test $ac_cv_func_syslog = no; then
       # syslog is not in the default libraries.  See if it's in some other.
       for lib in bsd socket inet; do
         AC_CHECK_LIB($lib, syslog, [AC_DEFINE(HAVE_SYSLOG)
           LIBS="$LIBS -l$lib"; break])

If you were working around bugs in AC_DEFINE_UNQUOTED by adding backslashes before quotes, you need to remove them. It now works predictably, and does not treat quotes (except back quotes) specially. See Setting Output Variables.

All of the Boolean shell variables set by Autoconf macros now use yes for the true value. Most of them use no for false, though for backward compatibility some use the empty string instead. If you were relying on a shell variable being set to something like 1 or t for true, you need to change your tests.

15.5.5 Changed Macro Writing

When defining your own macros, you should now use AC_DEFUN instead of define. AC_DEFUN automatically calls AC_PROVIDE and ensures that macros called via AC_REQUIRE do not interrupt other macros, to prevent nested checking... messages on the screen. There's no actual harm in continuing to use the older way, but it's less convenient and attractive. See Macro Definitions.

You probably looked at the macros that came with Autoconf as a guide for how to do things. It would be a good idea to take a look at the new versions of them, as the style is somewhat improved and they take advantage of some new features.

If you were doing tricky things with undocumented Autoconf internals (macros, variables, diversions), check whether you need to change anything to account for changes that have been made. Perhaps you can even use an officially supported technique in version 2 instead of kludging. Or perhaps not.

To speed up your locally written feature tests, add caching to them. See whether any of your tests are of general enough usefulness to encapsulate them into macros that you can share.

15.6 Upgrading From Version 2.13

The introduction of the previous section (see Autoconf 1) perfectly suits this section....

Autoconf version 2.50 is mostly backward compatible with version 2.13. However, it introduces better ways to do some things, and doesn't support some of the ugly things in version 2.13. So, depending on how sophisticated your files are, you might have to do some manual work in order to upgrade to version 2.50. This chapter points out some problems to watch for when upgrading. Also, perhaps your configure scripts could benefit from some of the new features in version 2.50; the changes are summarized in the file NEWS in the Autoconf distribution.

15.6.1 Changed Quotation

The most important changes are invisible to you: the implementation of most macros have completely changed. This allowed more factorization of the code, better error messages, a higher uniformity of the user's interface etc. Unfortunately, as a side effect, some construct which used to (miraculously) work might break starting with Autoconf 2.50. The most common culprit is bad quotation.

For instance, in the following example, the message is not properly quoted:

     AC_MSG_ERROR(cannot find foo.h, bailing out))

Autoconf 2.13 simply ignores it:

     $ autoconf-2.13; ./configure --silent
     creating cache ./config.cache
     configure: error: cannot find foo.h

while Autoconf 2.50 will produce a broken configure:

     $ autoconf-2.50; ./configure --silent
     configure: error: cannot find foo.h
     ./configure: exit: bad non-numeric arg `bailing'
     ./configure: exit: bad non-numeric arg `bailing'

The message needs to be quoted, and the AC_MSG_ERROR invocation too!

                      [AC_MSG_ERROR([cannot find foo.h, bailing out])])

Many many (and many more) Autoconf macros were lacking proper quotation, including no less than... AC_DEFUN itself!

     $ cat
     [# My own much better version
     $ autoconf-2.13
     autoconf: Undefined macros:
     ***BUG in Autoconf--please report*** AC_FD_MSG
     ***BUG in Autoconf--please report*** AC_EPI[AC_PROG_INSTALL],
     $ autoconf-2.50

15.6.2 New Macros

Because Autoconf has been dormant for years, Automake provided Autoconf-like macros for a while. Autoconf 2.50 now provides better versions of these macros, integrated in the AC_ namespace, instead of AM_. But in order to ease the upgrading via autoupdate, bindings to such AM_ macros are provided.

Unfortunately Automake did not quote the names of these macros! Therefore, when m4 finds something like AC_DEFUN(AM_TYPE_PTRDIFF_T, ...) in aclocal.m4, AM_TYPE_PTRDIFF_T is expanded, replaced with its Autoconf definition.

Fortunately Autoconf catches pre-AC_INIT expansions, and will complain, in its own words:

     $ cat
     $ aclocal-1.4
     $ autoconf
     ./aclocal.m4:17: error: m4_defn: undefined macro: _m4_divert_diversion
     actypes.m4:289: AM_TYPE_PTRDIFF_T is expanded from...
     ./aclocal.m4:17: the top level

Future versions of Automake will simply no longer define most of these macros, and will properly quote the names of the remaining macros. But you don't have to wait for it to happen to do the right thing right now: do not depend upon macros from Automake as it is simply not its job to provide macros (but the one it requires itself):

     $ cat
     $ rm aclocal.m4
     $ autoupdate
     autoupdate: `' is updated
     $ cat
     $ aclocal-1.4
     $ autoconf

15.6.3 Hosts and Cross-Compilation

Based on the experience of compiler writers, and after long public debates, many aspects of the cross-compilation chain have changed:

The relationship between build, host, and target have been cleaned up: the chain of default is now simply: target defaults to host, host to build, and build to the result of config.guess. Nevertheless, in order to ease the transition from 2.13 to 2.50, the following transition scheme is implemented. Do not rely on it, as it will be completely disabled in a couple of releases (we cannot keep it, as it proves to cause more problems than it cures).

They all default to the result of running config.guess, unless you specify either --build or --host. In this case, the default becomes the system type you specified. If you specify both, and they're different, configure will enter cross compilation mode, so it won't run any tests that require execution.

Hint: if you mean to override the result of config.guess, prefer --build over --host. In the future, --host will not override the name of the build system type. Whenever you specify --host, be sure to specify --build too.

For backward compatibility, configure will accept a system type as an option by itself. Such an option will override the defaults for build, host, and target system types. The following configure statement will configure a cross toolchain that will run on NetBSD/alpha but generate code for GNU Hurd/sparc, which is also the build platform.
     ./configure --host=alpha-netbsd sparc-gnu

In Autoconf 2.13 and before, the variables build, host, and target had a different semantics before and after the invocation of AC_CANONICAL_BUILD etc. Now, the argument of --build is strictly copied into build_alias, and is left empty otherwise. After the AC_CANONICAL_BUILD, build is set to the canonicalized build type. To ease the transition, before, its contents is the same as that of build_alias. Do not rely on this broken feature.

For consistency with the backward compatibility scheme exposed above, when --host is specified but --build isn't, the build system will be assumed to be the same as --host, and build_alias will be set to that value. Eventually, this historically incorrect behavior will go away.

The former scheme to enable cross-compilation proved to cause more harm than good, in particular, it used to be triggered too easily, leaving regular end users puzzled in front of cryptic error messages. configure could even enter cross-compilation mode only because the compiler was not functional. This is mainly because configure used to try to detect cross-compilation, instead of waiting for an explicit flag from the user.

Now, configure enters cross-compilation mode if and only if --host is passed.

That's the short documentation. To ease the transition between 2.13 and its successors, a more complicated scheme is implemented. Do not rely on the following, as it will be removed in the near future.

If you specify --host, but not --build, when configure performs the first compiler test it will try to run an executable produced by the compiler. If the execution fails, it will enter cross-compilation mode. This is fragile. Moreover, by the time the compiler test is performed, it may be too late to modify the build-system type: other tests may have already been performed. Therefore, whenever you specify --host, be sure to specify --build too.

     ./configure --build=i686-pc-linux-gnu --host=m68k-coff

will enter cross-compilation mode. The former interface, which consisted in setting the compiler to a cross-compiler without informing configure is obsolete. For instance, configure will fail if it can't run the code generated by the specified compiler if you configure as follows:

     ./configure CC=m68k-coff-gcc


Up to Autoconf 2.13, the replacement of functions was triggered via the variable LIBOBJS. Since Autoconf 2.50, the macro AC_LIBOBJ should be used instead (see Generic Functions). Starting at Autoconf 2.53, the use of LIBOBJS is an error.

This change is mandated by the unification of the GNU Build System components. In particular, the various fragile techniques used to parse a are all replaced with the use of traces. As a consequence, any action must be traceable, which obsoletes critical variable assignments. Fortunately, LIBOBJS was the only problem, and it can even be handled gracefully (read, “without your having to change something”).

There were two typical uses of LIBOBJS: asking for a replacement function, and adjusting LIBOBJS for Automake and/or Libtool.

As for function replacement, the fix is immediate: use AC_LIBOBJ. For instance:
     LIBOBJS="$LIBOBJS fnmatch.o"
     LIBOBJS="$LIBOBJS malloc.$ac_objext"

should be replaced with:


When asked for automatic de-ANSI-fication, Automake needs LIBOBJS'ed filenames to have $U appended to the base names. Libtool requires the definition of LTLIBOBJS, whose suffixes are mapped to .lo. People used to run snippets such as:
     # This is necessary so that .o files in LIBOBJS are also built via
     # the ANSI2KNR-filtering rules.
     LIBOBJS=`echo "$LIBOBJS" | sed 's/\.o /\$U.o /g;s/\.o$/\$U.o/'`
     LTLIBOBJS=`echo "$LIBOBJS" | sed 's/\.o/\.lo/g'`

Note that this code is wrong, because .o is not the only possible extension5! It should have read:

     # This is necessary so that .o files in LIBOBJS are also built via
     # the ANSI2KNR-filtering rules.
     LIB@&t@OBJS=`echo "$LIB@&t@OBJS" |
                  sed 's,\.[[^.]]* ,$U&,g;s,\.[[^.]]*$,$U&,'`
     LTLIBOBJS=`echo "$LIB@&t@OBJS" |
                sed 's,\.[[^.]]* ,.lo ,g;s,\.[[^.]]*$,.lo,'`

You no longer have to use this: AC_OUTPUT normalizes LIBOBJS and LTLIBOBJS (hence it works with any version of Automake and Libtool). Just remove these lines (autoupdate cannot handle this task, since this is not a macro).

Note that U must not be used in your Makefiles.


Since Autoconf 2.50, internal codes uses AC_PREPROC_IFELSE, AC_COMPILE_IFELSE, AC_LINK_IFELSE, and AC_RUN_IFELSE on one hand and AC_LANG_SOURCES, and AC_LANG_PROGRAM on the other hand instead of the deprecated AC_TRY_CPP, AC_TRY_COMPILE, AC_TRY_LINK, and AC_TRY_RUN. The motivations where:

In addition to the change of syntax, the philosphy has changed too: while emphasis was put on speed at the expense of accuracy, today's Autoconf promotes accuracy of the testing framework at, ahem..., the expense of speed.

As a perfect example of what is not to be done, here is how to find out whether a header file contains a particular declaration, such as a typedef, a structure, a structure member, or a function. Use AC_EGREP_HEADER instead of running grep directly on the header file; on some systems the symbol might be defined in another header file that the file you are checking #includes.

As a (bad) example, here is how you should not check for C preprocessor symbols, either defined by header files or predefined by the C preprocessor: using AC_EGREP_CPP:

     [#ifdef _AIX
     ], is_aix=yes, is_aix=no)

The above example, properly written would (i) use AC_LANG_PROGRAM, and (ii) run the compiler:

     [[#if !defined _AIX
     # error _AIX not defined

16 Generating Test Suites with Autotest

     N.B.: This section describes an experimental feature which will
     be part of Autoconf in a forthcoming release.  Although we believe
     Autotest is stabilizing, this documentation describes an interface which
     might change in the future: do not depend upon Autotest without
     subscribing to the Autoconf mailing lists.

It is paradoxical that portable projects depend on nonportable tools to run their test suite. Autoconf by itself is the paragon of this problem: although it aims at perfectly portability, up to 2.13, its test suite was using DejaGNU, a rich and complex testing framework, but which is far from being standard on Unix systems. Worse yet, it was likely to be missing on the most fragile platforms, the very platforms that are most likely to torture Autoconf and exhibit deficiencies.

To circumvent this problem many package maintainers have developed their own testing framework, based on simple shell scripts whose sole output are their exit status: the test succeeded, or failed. In addition, most of these tests share some common patterns, what results in lots of duplicated code, tedious maintenance etc.

Following exactly the same reasoning that yielded to the inception of Autoconf, Autotest provides a test suite generation frame work, based on M4 macros, building a portable shell script. The suite itself is equipped with automatic logging and tracing facilities which greatly diminish the interaction with bug reporters, and simple timing reports.

Autoconf itself has been using Autotest for years, and we do attest that it has considerably improved the strength of the test suite, and the quality of bug reports. Other projects are known to use some generation of Autotest, such as Bison, Free Recode, Free Wdiff, GNU Tar, each of them having different needs, what slowly polishes Autotest as a general testing framework.

Nonetheless, compared to DejaGNU, Autotest is inadequate for interactive tool testing, which is probably its main limitation.

16.1 Using an Autotest Test Suite

16.1.1 testsuite Scripts

Generating testing or validation suites using Autotest is rather easy. The whole validation suite is held in a file to be processed through autom4te, itself using GNU M4 under the scene, to produce a stand-alone Bourne shell script which then gets distributed. Neither autom4te nor GNU M4 are not needed anymore at the installer end.

Each test of the validation suite should be part of some test group. A test group is a sequence of interwoven tests that ought to be executed together, usually because one test in the group creates data files than a later test in the same group needs to read. Complex test groups make later debugging more tedious. It is much better keeping keep only a few tests per test group, and if you can put only one test per test group, this is just ideal.

For all but the simplest packages, some file such as does not fully hold all test sources, as these are often easier to maintain in separate files. Each of these separate files holds a single test group, or a sequence of test groups all addressing some common functionality in the package. In such cases, file only initializes the whole validation suite, and sometimes do elementary health checking, before listing include statements for all other test files. The special file package.m4, containing the identification of the package, is automatically included if found.

A convenient alternative consists in moving all the global issues (local Autotest macros, elementary health checking, and AT_INIT invocation) into the file, and making be a simple list of m4_include of sub test suites. In such case, generating the whole test suite or pieces of it is only a matter of choosing the autom4te command line arguments.

The validation scripts that Autotest produces are by convention called testsuite. When run, testsuite executes each test group in turn, producing only one summary line per test to say if that particular test succeeded or failed. At end of all tests, summarizing counters get printed. One debugging directory is left for each test group which failed, if any: such directories are named testsuite.dir/nn, where nn is the sequence number of the test group, and they include:

In the ideal situation, none of the tests fail, and consequently, no debugging directory is left out of validation.

It often happens in practice that individual tests in the validation suite need to get information coming out of the configuration process. Some of this information, common for all validation suites, is provided through the file atconfig, automatically created by AC_CONFIG_TESTDIR. For configuration informations which your testing environment specifically needs, you might prepare an optional file named, instantiated by AC_CONFIG_FILES. The configuration process produces atconfig and atlocal out of these two input files, and these two produced files are automatically read by the testsuite script.

Here is a diagram showing the relationship between files.

Files used in preparing a software package for distribution:

                     [package.m4] -->.
                                      \ ->.  [] ---->+
         ...         \                  \ ---->-- -->-- autom4te* -->testsuite
         ...         / ->'

Files used in configuring a software package:

                                          .--> atconfig
     [] -->  config.status* --<
                                          `--> [atlocal]

Files created during the test suite execution:

     atconfig -->.                    .--> testsuite.log
                  \                  /
                   >-- testsuite* --<
                  /                  \
     [atlocal] ->'                    `--> [testsuite.dir]

16.1.2 Autotest Logs

When run, the test suite creates a log file named after itself, e.g., a test suite named testsuite creates testsuite.log. It contains a lot of information, usually more than maintainers actually need, but therefore most of the time it contains all that is needed:

command line arguments
A very bad Unix habit which is unfortunately wide spread consists of setting environment variables before the command, such as in CC=my-home-grown-cc ./testsuite. This results in the test suite not knowing this change, hence (i) it can't report it to you, and (ii) it cannot preserve the value of CC for subsequent runs. Autoconf faced exactly the same problem, and solved it by asking users to pass the variable definitions as command line arguments. Autotest requires this rule too, but has no means to enforce it; the log then contains a trace of the variables the user changed.
ChangeLog excerpts
The topmost lines of all the ChangeLogs found in the source hierarchy. This is especially useful when bugs are reported against development versions of the package, since the version string does not provide sufficient information to know the exact state of the sources the user compiled. Of course this relies on the use of a ChangeLog.
build machine
Running a test suite in a cross-compile environment is not an easy task, since it would mean having the test suite run on a machine build, while running programs on a machine host. It is much simpler to run both the test suite and the programs on host, but then, from the point of view of the test suite, there remains a single environment, host = build. The log contains relevant information on the state of the build machine, including some important environment variables.
tested programs
The absolute path and answers to --version of the tested programs (see Writing, AT_TESTED).
configuration log
The contents of config.log, as created by configure, are appended. It contains the configuration flags and a detailed report on the configuration itself.

16.2 Writing

The is a Bourne shell script making use of special Autotest M4 macros. It often contains a call to AT_INIT nears its beginning followed by one call to m4_include per source file for tests. Each such included file, or the remainder of if include files are not used, contain a sequence of test groups. Each test group begins with one call to AT_SETUP, it contains an arbitrary number of shell commands or calls to AT_CHECK, and it completes with one call to AT_CLEANUP.

— Macro: AT_INIT ([name])

Initialize Autotest. Giving a name to the test suite is encouraged if your package includes several test suites. In any case, the test suite always displays the package name and version. It also inherits the package bug report address.

— Macro: AT_TESTED (executables)

Log the path and answer to --version of each program in space-separated list executables. Several invocations register new executables, in other words, don't fear registering one program several times.

Autotest test suites rely on the PATH to find the tested program. This saves from generating the absolute paths to the various tools, and makes it possible to test installed programs. Therefore, knowing what programs are being exercised is crucial to understand some problems in the test suite itself, or its occasional misuses. It is a good idea to also subscribe foreign programs you depend upon, to ease incompatibility diagnostics.

— Macro: AT_SETUP (test-group-name)

This macro starts a group of related tests, all to be executed in the same subshell. It accepts a single argument, which holds a few words (no more than about 30 or 40 characters) quickly describing the purpose of the test group being started.

— Macro: AT_KEYWORDS (keywords)

Associate the space-separated list of keywords to the enclosing test group. This makes it possible to run “slices” of the test suite. For instance if some of your test groups exercise some foo feature, then using AT_KEYWORDS(foo) lets you run ./testsuite -k foo to run exclusively these test groups. The title of the test group is automatically recorded to AT_KEYWORDS.

Several invocations within a test group accumulate new keywords. In other words, don't fear registering several times the same keyword in a test group.

— Macro: AT_XFAIL_IF (shell-condition)

Determine whether the test is expected to fail because it is a known bug (for unsupported features, you should skip the test). shell-condition is a shell expression such as a test command; you can instantiate this macro many times from within the same test group, and one of the conditions will be enough to turn the test into an expected failure.


End the current test group.

— Macro: AT_DATA (file, contents)

Initialize an input data file with given contents. Of course, the contents have to be properly quoted between square brackets to protect against included commas or spurious M4 expansion. The contents ought to end with an end of line.

— Macro: AT_CHECK (commands, [status = 0], [stdout = ], [stderr = ], [run-if-fail], [run-if-pass])

Execute a test by performing given shell commands. These commands should normally exit with status, while producing expected stdout and stderr contents. If commands exit with status 77, then the whole test group is skipped. Otherwise, if this test fails, run shell commands run-if-fail or, if this test passes, run shell commands run-if-pass.

The commands must not redirect the standard output, nor the standard error.

If status, or stdout, or stderr is ignore, then the corresponding value is not checked.

The special value expout for stdout means the expected output of the commands is the content of the file expout. If stdout is stdout, then the standard output of the commands is available for further tests in the file stdout. Similarly for stderr with expout and stderr.

16.3 Running testsuite Scripts

Autotest test suites support the following arguments:

Display the list of options and exit successfully.
Display the version of the test suite and exit successfully.
Remove all the files the test suite might have created and exit. Meant for clean Makefile targets.
List all the tests (or only the selection), including their possible keywords.

By default all the tests are performed (or described with --list) in the default environment first silently, then verbosely, but the environment, set of tests, and verbosity level can be tuned:
Set the environment variable to value. Do not run FOO=foo ./testsuite as debugging scripts would then run in a different environment.

The variable AUTOTEST_PATH specifies the testing path to prepend to PATH. It handles specially relative paths (not starting with /): they are considered to be relative to the top level of the package being built. All the directories are made absolute, first starting from the top level build tree, then from the source tree. For instance ./testsuite AUTOTEST_PATH=tests:bin for a /src/foo-1.0 source package built in /tmp/foo results in /tmp/foo/tests:/tmp/foo/bin and then /src/foo-1.0/tests:/src/foo-1.0/bin being prepended to PATH.

Add the corresponding test groups, with obvious semantics, to the selection.
-k keywords
Add to the selection the test groups which title or keywords (arguments to AT_SETUP or AT_KEYWORDS) match all the keywords of the comma separated list keywords.

Running ./testsuite -k autoupdate,FUNC will select all the tests tagged with autoupdate and FUNC (as in AC_CHECK_FUNC, AC_FUNC_FNMATCH etc.) while ./testsuite -k autoupdate -k FUNC runs all the tests tagged with autoupdate or FUNC.

If any test fails, immediately abort testing. It implies --debug: post test group clean up, debugging script generation, and logging are inhibited. This option is meant for the full test suite, it is not really useful for generated debugging scripts.
Force more verbosity in the detailed output of what is being done. This is the default for debugging scripts.
Do not remove the files after a test group was performed —but they are still removed before, therefore using this option is sane when running several test groups. Do not create debugging scripts. Do not log (in order to preserve supposedly existing full log file). This is the default for debugging scripts, but it can also be useful to debug the testsuite itself.
Trigger shell tracing of the test groups.

16.4 Making testsuite Scripts

For putting Autotest into movement, you need some configuration and Makefile machinery. We recommend, at least if your package uses deep or shallow hierarchies, that you use tests/ as the name of the directory holding all your tests and their Makefile. Here is a check list of things to do.

With Automake, here is a minimal example about how to link make check with a validation suite.

     EXTRA_DIST = testsuite
     TESTSUITE = $(srcdir)/testsuite
     check-local: atconfig atlocal $(TESTSUITE)
             $(SHELL) $(TESTSUITE)
     AUTOTEST = $(AUTOM4TE) --language=autotest
     $(TESTSUITE): $(srcdir)/
             $(AUTOTEST) -I $(srcdir) $ -o $@.tmp
             mv $@.tmp $@

You might want to list explicitly the dependencies, i.e., the list of the files includes.

With strict Autoconf, you might need to add lines inspired from the following:

     subdir = tests
     atconfig: $(top_builddir)/config.status
             cd $(top_builddir) && \
                $(SHELL) ./config.status $(subdir)/$@
     atlocal: $(srcdir)/ $(top_builddir)/config.status
             cd $(top_builddir) && \
                $(SHELL) ./config.status $(subdir)/$@

and manage to have and $(EXTRA_DIST) distributed.

17 Frequent Autoconf Questions, with answers

Several questions about Autoconf come up occasionally. Here some of them are addressed.

17.1 Distributing configure Scripts

     What are the restrictions on distributing configure
     scripts that Autoconf generates?  How does that affect my
     programs that use them?

There are no restrictions on how the configuration scripts that Autoconf produces may be distributed or used. In Autoconf version 1, they were covered by the GNU General Public License. We still encourage software authors to distribute their work under terms like those of the GPL, but doing so is not required to use Autoconf.

Of the other files that might be used with configure, is under whatever copyright you use for your config.sub and config.guess have an exception to the GPL when they are used with an Autoconf-generated configure script, which permits you to distribute them under the same terms as the rest of your package. install-sh is from the X Consortium and is not copyrighted.

17.2 Why Require GNU M4?

     Why does Autoconf require GNU M4?

Many M4 implementations have hard-coded limitations on the size and number of macros that Autoconf exceeds. They also lack several builtin macros that it would be difficult to get along without in a sophisticated application like Autoconf, including:


Autoconf requires version 1.4 or above of GNU M4 because it uses frozen state files.

Since only software maintainers need to use Autoconf, and since GNU M4 is simple to configure and install, it seems reasonable to require GNU M4 to be installed also. Many maintainers of GNU and other free software already have most of the GNU utilities installed, since they prefer them.

17.3 How Can I Bootstrap?

     If Autoconf requires GNU M4 and GNU M4 has an Autoconf
     configure script, how do I bootstrap?  It seems like a chicken
     and egg problem!

This is a misunderstanding. Although GNU M4 does come with a configure script produced by Autoconf, Autoconf is not required in order to run the script and install GNU M4. Autoconf is only required if you want to change the M4 configure script, which few people have to do (mainly its maintainer).

17.4 Why Not Imake?

     Why not use Imake instead of configure scripts?

Several people have written addressing this question, so I include adaptations of their explanations here.

The following answer is based on one written by Richard Pixley:

Autoconf generated scripts frequently work on machines that it has never been set up to handle before. That is, it does a good job of inferring a configuration for a new system. Imake cannot do this.

Imake uses a common database of host specific data. For X11, this makes sense because the distribution is made as a collection of tools, by one central authority who has control over the database.

GNU tools are not released this way. Each GNU tool has a maintainer; these maintainers are scattered across the world. Using a common database would be a maintenance nightmare. Autoconf may appear to be this kind of database, but in fact it is not. Instead of listing host dependencies, it lists program requirements.

If you view the GNU suite as a collection of native tools, then the problems are similar. But the GNU development tools can be configured as cross tools in almost any host+target permutation. All of these configurations can be installed concurrently. They can even be configured to share host independent files across hosts. Imake doesn't address these issues.

Imake templates are a form of standardization. The GNU coding standards address the same issues without necessarily imposing the same restrictions.

Here is some further explanation, written by Per Bothner:

One of the advantages of Imake is that it easy to generate large Makefiles using cpp's #include and macro mechanisms. However, cpp is not programmable: it has limited conditional facilities, and no looping. And cpp cannot inspect its environment.

All of these problems are solved by using sh instead of cpp. The shell is fully programmable, has macro substitution, can execute (or source) other shell scripts, and can inspect its environment.

Paul Eggert elaborates more:

With Autoconf, installers need not assume that Imake itself is already installed and working well. This may not seem like much of an advantage to people who are accustomed to Imake. But on many hosts Imake is not installed or the default installation is not working well, and requiring Imake to install a package hinders the acceptance of that package on those hosts. For example, the Imake template and configuration files might not be installed properly on a host, or the Imake build procedure might wrongly assume that all source files are in one big directory tree, or the Imake configuration might assume one compiler whereas the package or the installer needs to use another, or there might be a version mismatch between the Imake expected by the package and the Imake supported by the host. These problems are much rarer with Autoconf, where each package comes with its own independent configuration processor.

Also, Imake often suffers from unexpected interactions between make and the installer's C preprocessor. The fundamental problem here is that the C preprocessor was designed to preprocess C programs, not Makefiles. This is much less of a problem with Autoconf, which uses the general-purpose preprocessor M4, and where the package's author (rather than the installer) does the preprocessing in a standard way.

Finally, Mark Eichin notes:

Imake isn't all that extensible, either. In order to add new features to Imake, you need to provide your own project template, and duplicate most of the features of the existing one. This means that for a sophisticated project, using the vendor-provided Imake templates fails to provide any leverage—since they don't cover anything that your own project needs (unless it is an X11 program).

On the other side, though:

The one advantage that Imake has over configure: Imakefiles tend to be much shorter (likewise, less redundant) than Makefile.ins. There is a fix to this, however—at least for the Kerberos V5 tree, we've modified things to call in common and Makefile fragments for the entire tree. This means that a lot of common things don't have to be duplicated, even though they normally are in configure setups.

17.5 How Do I #define Installation Directories?

     My program needs library files, installed in datadir and
     similar.  If I use
          AC_DEFINE_UNQUOTED([DATADIR], [$datadir],
                             [Define to the read-only architecture-independent
                              data directory.])

I get

          #define DATADIR "${prefix}/share"

As already explained, this behavior is on purpose, mandated by the GNU Coding Standards, see Installation Directory Variables. There are several means to achieve a similar goal:

17.6 What is autom4te.cache?

     What is this directory autom4te.cache?  Can I safely remove it?

In the GNU Build System, plays a central role and is read by many tools: autoconf to create configure, autoheader to create, automake to create, autoscan to check the completeness of, autoreconf to check the GNU Build System components that are used. To “read” actually means to compile it with M4, which can be a very long process for complex

This is why all these tools, instead of running directly M4, invoke autom4te (see autom4te Invocation) which, while answering to a specific demand, stores additional information in autom4te.cache for future runs. For instance, if you run autoconf, behind the scenes, autom4te will also store information for the other tools, so that when you invoke autoheader or automake etc., re-processing is not needed. The speed up is frequently of 30, and is increasing with the size of

But it is and remains being simply a cache: you can safely remove it.

     Can I permanently get rid of it?

The creation of this cache can be disabled from ~/.autom4te.cfg, see Customizing autom4te, for more details. You should be aware that disabling the cache slows down the Autoconf test suite by 40%. The more GNU Build System components are used, the more the cache is useful; for instance running autoreconf -f on the Coreutils is twice slower without the cache although --force implies that the cache is not fully exploited, and eight times slower than without --force.

17.7 Header Present But Cannot Be Compiled

The most important guideline to bear in mind when checking for features is to mimic as much as possible the intended use. Unfortunately, old versions of AC_CHECK_HEADER and AC_CHECK_HEADERS failed to follow this idea, and called the preprocessor, instead of the compiler, to check for headers. As a result, incompatibilities between headers went unnoticed during configuration, and maintainers finally had to deal with this issue elsewhere.

As of Autoconf 2.56 both checks are performed, and configure complains loudly if the compiler and the preprocessor do not agree. For the time being the result used is that of the preprocessor, to give maintainers time to adjust their, but in the near future, only the compiler will be considered.

Consider the following example:

     $ cat number.h
     typedef int number;
     $ cat pi.h
     const number pi = 3;
     $ cat
     $ autoconf -Wall
     $ ./configure
     checking for gcc... gcc
     checking for C compiler default output... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ANSI C... none needed
     checking how to run the C preprocessor... gcc -E
     checking for egrep... grep -E
     checking for ANSI C header files... yes
     checking for sys/types.h... yes
     checking for sys/stat.h... yes
     checking for stdlib.h... yes
     checking for string.h... yes
     checking for memory.h... yes
     checking for strings.h... yes
     checking for inttypes.h... yes
     checking for stdint.h... yes
     checking for unistd.h... yes
     checking pi.h usability... no
     checking pi.h presence... yes
     configure: WARNING: pi.h: present but cannot be compiled
     configure: WARNING: pi.h: check for missing prerequisite headers?
     configure: WARNING: pi.h: proceeding with the preprocessor's result
     configure: WARNING:     ## ------------------------------------ ##
     configure: WARNING:     ## Report this to ##
     configure: WARNING:     ## ------------------------------------ ##
     checking for pi.h... yes

The proper way the handle this case is using the fourth argument (see Generic Headers):

     $ cat
     AC_CHECK_HEADERS(number.h pi.h,,,
     [[#if HAVE_NUMBER_H
     # include <number.h>
     $ autoconf -Wall
     $ ./configure
     checking for gcc... gcc
     checking for C compiler default output... a.out
     checking whether the C compiler works... yes
     checking whether we are cross compiling... no
     checking for suffix of executables...
     checking for suffix of object files... o
     checking whether we are using the GNU C compiler... yes
     checking whether gcc accepts -g... yes
     checking for gcc option to accept ANSI C... none needed
     checking for number.h... yes
     checking for pi.h... yes

See Particular Headers, for a list of headers with their prerequisite.

18 History of Autoconf

You may be wondering, Why was Autoconf originally written? How did it get into its present form? (Why does it look like gorilla spit?) If you're not wondering, then this chapter contains no information useful to you, and you might as well skip it. If you are wondering, then let there be light....

18.1 Genesis

In June 1991 I was maintaining many of the GNU utilities for the Free Software Foundation. As they were ported to more platforms and more programs were added, the number of -D options that users had to select in the Makefile (around 20) became burdensome. Especially for me—I had to test each new release on a bunch of different systems. So I wrote a little shell script to guess some of the correct settings for the fileutils package, and released it as part of fileutils 2.0. That configure script worked well enough that the next month I adapted it (by hand) to create similar configure scripts for several other GNU utilities packages. Brian Berliner also adapted one of my scripts for his CVS revision control system.

Later that summer, I learned that Richard Stallman and Richard Pixley were developing similar scripts to use in the GNU compiler tools; so I adapted my configure scripts to support their evolving interface: using the file name as the templates; adding +srcdir, the first option (of many); and creating config.status files.

18.2 Exodus

As I got feedback from users, I incorporated many improvements, using Emacs to search and replace, cut and paste, similar changes in each of the scripts. As I adapted more GNU utilities packages to use configure scripts, updating them all by hand became impractical. Rich Murphey, the maintainer of the GNU graphics utilities, sent me mail saying that the configure scripts were great, and asking if I had a tool for generating them that I could send him. No, I thought, but I should! So I started to work out how to generate them. And the journey from the slavery of hand-written configure scripts to the abundance and ease of Autoconf began.

Cygnus configure, which was being developed at around that time, is table driven; it is meant to deal mainly with a discrete number of system types with a small number of mainly unguessable features (such as details of the object file format). The automatic configuration system that Brian Fox had developed for Bash takes a similar approach. For general use, it seems to me a hopeless cause to try to maintain an up-to-date database of which features each variant of each operating system has. It's easier and more reliable to check for most features on the fly—especially on hybrid systems that people have hacked on locally or that have patches from vendors installed.

I considered using an architecture similar to that of Cygnus configure, where there is a single configure script that reads pieces of when run. But I didn't want to have to distribute all of the feature tests with every package, so I settled on having a different configure made from each by a preprocessor. That approach also offered more control and flexibility.

I looked briefly into using the Metaconfig package, by Larry Wall, Harlan Stenn, and Raphael Manfredi, but I decided not to for several reasons. The Configure scripts it produces are interactive, which I find quite inconvenient; I didn't like the ways it checked for some features (such as library functions); I didn't know that it was still being maintained, and the Configure scripts I had seen didn't work on many modern systems (such as System V R4 and NeXT); it wasn't very flexible in what it could do in response to a feature's presence or absence; I found it confusing to learn; and it was too big and complex for my needs (I didn't realize then how much Autoconf would eventually have to grow).

I considered using Perl to generate my style of configure scripts, but decided that M4 was better suited to the job of simple textual substitutions: it gets in the way less, because output is implicit. Plus, everyone already has it. (Initially I didn't rely on the GNU extensions to M4.) Also, some of my friends at the University of Maryland had recently been putting M4 front ends on several programs, including tvtwm, and I was interested in trying out a new language.

18.3 Leviticus

Since my configure scripts determine the system's capabilities automatically, with no interactive user intervention, I decided to call the program that generates them Autoconfig. But with a version number tacked on, that name would be too long for old unix file systems, so I shortened it to Autoconf.

In the fall of 1991 I called together a group of fellow questers after the Holy Grail of portability (er, that is, alpha testers) to give me feedback as I encapsulated pieces of my handwritten scripts in M4 macros and continued to add features and improve the techniques used in the checks. Prominent among the testers were François Pinard, who came up with the idea of making an Autoconf shell script to run M4 and check for unresolved macro calls; Richard Pixley, who suggested running the compiler instead of searching the file system to find include files and symbols, for more accurate results; Karl Berry, who got Autoconf to configure TeX and added the macro index to the documentation; and Ian Lance Taylor, who added support for creating a C header file as an alternative to putting -D options in a Makefile, so he could use Autoconf for his UUCP package. The alpha testers cheerfully adjusted their files again and again as the names and calling conventions of the Autoconf macros changed from release to release. They all contributed many specific checks, great ideas, and bug fixes.

18.4 Numbers

In July 1992, after months of alpha testing, I released Autoconf 1.0, and converted many GNU packages to use it. I was surprised by how positive the reaction to it was. More people started using it than I could keep track of, including people working on software that wasn't part of the GNU Project (such as TCL, FSP, and Kerberos V5). Autoconf continued to improve rapidly, as many people using the configure scripts reported problems they encountered.

Autoconf turned out to be a good torture test for M4 implementations. unix M4 started to dump core because of the length of the macros that Autoconf defined, and several bugs showed up in GNU M4 as well. Eventually, we realized that we needed to use some features that only GNU M4 has. 4.3BSD M4, in particular, has an impoverished set of builtin macros; the System V version is better, but still doesn't provide everything we need.

More development occurred as people put Autoconf under more stresses (and to uses I hadn't anticipated). Karl Berry added checks for X11. david zuhn contributed C++ support. François Pinard made it diagnose invalid arguments. Jim Blandy bravely coerced it into configuring GNU Emacs, laying the groundwork for several later improvements. Roland McGrath got it to configure the GNU C Library, wrote the autoheader script to automate the creation of C header file templates, and added a --verbose option to configure. Noah Friedman added the --autoconf-dir option and AC_MACRODIR environment variable. (He also coined the term autoconfiscate to mean “adapt a software package to use Autoconf”.) Roland and Noah improved the quoting protection in AC_DEFINE and fixed many bugs, especially when I got sick of dealing with portability problems from February through June, 1993.

18.5 Deuteronomy

A long wish list for major features had accumulated, and the effect of several years of patching by various people had left some residual cruft. In April 1994, while working for Cygnus Support, I began a major revision of Autoconf. I added most of the features of the Cygnus configure that Autoconf had lacked, largely by adapting the relevant parts of Cygnus configure with the help of david zuhn and Ken Raeburn. These features include support for using config.sub, config.guess, --host, and --target; making links to files; and running configure scripts in subdirectories. Adding these features enabled Ken to convert GNU as, and Rob Savoye to convert DejaGNU, to using Autoconf.

I added more features in response to other peoples' requests. Many people had asked for configure scripts to share the results of the checks between runs, because (particularly when configuring a large source tree, like Cygnus does) they were frustratingly slow. Mike Haertel suggested adding site-specific initialization scripts. People distributing software that had to unpack on MS-DOS asked for a way to override the .in extension on the file names, which produced file names like containing two dots. Jim Avera did an extensive examination of the problems with quoting in AC_DEFINE and AC_SUBST; his insights led to significant improvements. Richard Stallman asked that compiler output be sent to config.log instead of /dev/null, to help people debug the Emacs configure script.

I made some other changes because of my dissatisfaction with the quality of the program. I made the messages showing results of the checks less ambiguous, always printing a result. I regularized the names of the macros and cleaned up coding style inconsistencies. I added some auxiliary utilities that I had developed to help convert source code packages to use Autoconf. With the help of François Pinard, I made the macros not interrupt each others' messages. (That feature revealed some performance bottlenecks in GNU M4, which he hastily corrected!) I reorganized the documentation around problems people want to solve. And I began a test suite, because experience had shown that Autoconf has a pronounced tendency to regress when we change it.

Again, several alpha testers gave invaluable feedback, especially François Pinard, Jim Meyering, Karl Berry, Rob Savoye, Ken Raeburn, and Mark Eichin.

Finally, version 2.0 was ready. And there was much rejoicing. (And I have free time again. I think. Yeah, right.)

Appendix A Copying This Manual

A.1 GNU Free Documentation License

Version 1.2, November 2002
     Copyright © 2000,2001,2002 Free Software Foundation, Inc.
     59 Temple Place, Suite 330, Boston, MA  02111-1307, USA
     Everyone is permitted to copy and distribute verbatim copies
     of this license document, but changing it is not allowed.

    The purpose of this License is to make a manual, textbook, or other functional and useful document free in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others.

    This License is a kind of “copyleft”, which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software.

    We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.


    This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that work under the conditions stated herein. The “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept the license if you copy, modify or distribute the work in a way requiring permission under copyright law.

    A “Modified Version” of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language.

    A “Secondary Section” is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them.

    The “Invariant Sections” are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none.

    The “Cover Texts” are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most 25 words.

    A “Transparent” copy of the Document means a machine-readable copy, represented in a format whose specification is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any substantial amount of text. A copy that is not “Transparent” is called “Opaque”.

    Examples of suitable formats for Transparent copies include plain ascii without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only.

    The “Title Page” means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, “Title Page” means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text.

    A section “Entitled XYZ” means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name mentioned below, such as “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according to this definition.

    The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License.


    You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3.

    You may also lend copies, under the same conditions stated above, and you may publicly display copies.


    If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects.

    If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages.

    If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public.

    It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.


    You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version:

    1. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission.
    2. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has fewer than five), unless they release you from this requirement.
    3. State on the Title page the name of the publisher of the Modified Version, as the publisher.
    4. Preserve all the copyright notices of the Document.
    5. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices.
    6. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below.
    7. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document's license notice.
    8. Include an unaltered copy of this License.
    9. Preserve the section Entitled “History”, Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled “History” in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence.
    10. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the “History” section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission.
    11. For any section Entitled “Acknowledgements” or “Dedications”, Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein.
    12. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles.
    13. Delete any section Entitled “Endorsements”. Such a section may not be included in the Modified Version.
    14. Do not retitle any existing section to be Entitled “Endorsements” or to conflict in title with any Invariant Section.
    15. Preserve any Warranty Disclaimers.

    If the Modified Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles.

    You may add a section Entitled “Endorsements”, provided it contains nothing but endorsements of your Modified Version by various parties—for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard.

    You may add a passage of up to five words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one.

    The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modified Version.


    You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers.

    The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work.

    In the combination, you must combine any sections Entitled “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled “Dedications”. You must delete all sections Entitled “Endorsements.”


    You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects.

    You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document.


    A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an “aggregate” if the copyright resulting from the compilation is not used to limit the legal rights of the compilation's users beyond what the individual works permit. When the Document is included an aggregate, this License does not apply to the other works in the aggregate which are not themselves derivative works of the Document.

    If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Document's Cover Texts may be placed on covers that bracket the Document within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate.


    Translation is considered a kind of modification, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warrany Disclaimers, provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer, the original version will prevail.

    If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title.


    You may not copy, modify, sublicense, or distribute the Document except as expressly provided for under this License. Any other attempt to copy, modify, sublicense or distribute the Document is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance.


    The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. See

    Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License “or any later version” applies to it, you have the option of following the terms and conditions either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation.

A.1.1 ADDENDUM: How to use this License for your documents

To use this License in a document you have written, include a copy of the License in the document and put the following copyright and license notices just after the title page:

       Copyright (C)  year  your name.
       Permission is granted to copy, distribute and/or modify this document
       under the terms of the GNU Free Documentation License, Version 1.2
       or any later version published by the Free Software Foundation;
       with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
       A copy of the license is included in the section entitled ``GNU
       Free Documentation License''.

If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the “with...Texts.” line with this:

         with the Invariant Sections being list their titles, with
         the Front-Cover Texts being list, and with the Back-Cover Texts
         being list.

If you have Invariant Sections without Cover Texts, or some other combination of the three, merge those two alternatives to suit the situation.

If your document contains nontrivial examples of program code, we recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software.

Appendix B Indices

B.1 Environment Variable Index

This is an alphabetical list of the environment variables that Autoconf checks.

B.2 Output Variable Index

This is an alphabetical list of the variables that Autoconf can substitute into files that it creates, typically one or more Makefiles. See Setting Output Variables, for more information on how this is done.

B.3 Preprocessor Symbol Index

This is an alphabetical list of the C preprocessor symbols that the Autoconf macros define. To work with Autoconf, C source code needs to use these names in #if directives.

B.4 Autoconf Macro Index

This is an alphabetical list of the Autoconf macros.

B.5 M4 Macro Index

This is an alphabetical list of the M4, M4sugar, and M4sh macros.

B.6 Autotest Macro Index

This is an alphabetical list of the Autotest macros.

B.7 Program and Function Index

This is an alphabetical list of the programs and functions which portability is discussed in this document.

B.8 Concept Index

This is an alphabetical list of the files, tools, and concepts introduced in this document.


[1] GNU Autoconf, Automake and Libtool, by G. V. Vaughan, B. Elliston, T. Tromey, and I. L. Taylor. New Riders, 2000, ISBN 1578701902.

[2] Because M4 is not aware of Sh code, especially conditionals, some optimizations that look nice statically may produce incorrect results at runtime.

[3] Using defn.

[4] Yet another great name from Lars J. Aas.

[5] Yet another reason why assigning LIBOBJS directly is discouraged.