* doc/autoconf.texi (C and Posix Variants):
Don’t insist on C11 when talking about Annex K, as
Annex K is also present in later versions of the C standard.
A lot of bugs recently have been due to unexpected changes in shell
and/or make behavior, so let’s log their versions in testsuite.log
to make it easier to figure out when that’s the problem.
Some shells don’t recognize --version but we still get an indication
in the logfile of which shell it is, e.g.
$ make check TESTSUITEFLAGS=1 SHELL=/bin/dash; grep dash tests/testsuite.log
../../tests/local.at:25: /bin/dash --version
/bin/dash: 0: Illegal option --
* tests/local.at: Add ${CONFIG_SHELL-$SHELL} and $MAKE to AT_TESTED.
Revise all NEWS entries since 2.71 for clarity. Add an entry
to the compatibility section about the macros that no longer
have AC_PROG_{EGREP,CPP} as a side effect.
* lib/m4sugar/m4sugar.m4 (m4_file_append): Remove.
In 2000, this undocumented macro was left in by mistake
even though all its callers were removed.
Inspired by a bug report by Ally Petitt (sr#110961).
AS_ECHO’s argument is required to be “a single shell word,”
and this more precisely means it must not be altered by shell
word splitting. In particular, if the argument contains shell
variables whose values contain whitespace then it needs to be
wrapped in "shell double quotes".
The absence of these quotes caused one of the embedded M4 scripts
to be mangled by the Autoconf 2.69 implementation of AS_ECHO.
We don’t officially support bootstrapping with an older version
of Autoconf (use the ./bootstrap script instead) but the absence
of quotes is still incorrect.
For consistency add [M4 quotes] to the use of AS_ECHO that was
shell-quoted but not M4-quoted.
* m4/m4.m4 (AC_PROG_GNU_M4): Quote argument to AS_ECHO correctly.
On OS/2, chmod(2) cannot be applied to an open file.
Instead set the desired permissions when the file is initially
created, using the PERMS argument to File::Temp::tempfile if
possible, or by manually emulating that feature if the system
perl does not provide a new enough version of File::Temp.
This has the nice side effect that we no longer need to handle
the umask manually.
* autom4te.in (tempfile_with_mode): New function.
(handle_output): Use tempfile_with_mode instead of directly using
File::Temp plus chmod.
Co-authored-by: KO Myung-Hun <komh78@gmail.com>
Following the glibc commit glibc-2.38~268 ("libio: Add __nonnull
for FILE * arguments of fclose and freopen"), the utilization of the gcc
-fanalyzer flag in combination with the -Werror flag leads to an error
within the _AC_LANG_IO_PROGRAM(C) function:
conftest.c:14:23: error: use of possibly-NULL 'f' where non-null expected [CWE-690] [-Werror=analyzer-possible-null-argument]
* lib/autoconf/c.m4 (_AC_LANG_IO_PROGRAM): Exit upon encountering
fopen() error.
See also https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109570 and
https://gcc.gnu.org/g:c9a19621a07e246385ae075b61283140b23c3b5a .
Tiny-change: yes
Rewrite install.texi so that INSTALL is shorter and hopefully
more useful for new installers.
* doc/install.texi: When not building the Autoconf manual, put
copyright notice at the end, and omit external references as they
do not work in INSTALL.
Avoid long URLs as they do not work well in INSTALL or
in info files. Be more consistent (though not entirely
consistent) about “system” vs “architecture”.
(System Types, Building for a different system type)
(Troubleshooting the Build Type): Coalesce into a single section
and rewrite for clarity and brevity. Mention build-aux. Do not
mention --target here as it’s too rare to make the cut.
* doc/install.texi (Compilers and Options): Add another reference.
(System Types): Renamed from System Type. Explain how to canonicalize
and how to validate a system type. Don't explain --build, --host,
--target here.
(Building for a different system type): New section.
(Troubleshooting the Build Type): New section.
(Configuring a Compiler): New section.
(configure Invocation): Mention the --host option, not the --build
option, since --build is so rarely needed.
Problem reported for gettext bootstrap by Julien Palard in:
https://savannah.gnu.org/bugs/?62196
* doc/install.texi: Give a brief info as to how to bootstrap,
on packages built from Git rather from a distribution tarball.
Remove aging details about c99, macOS, X, HP-UX, OSF/1, Solaris,
Haiku, sun4. Improve documentation for enable/disable and
with/without options.
Problem reported by Matoro <https://bugs.gnu.org/64123>.
* lib/autoconf/specific.m4 (_AC_SYS_LARGEFILE_TEST_CODE):
New overridable macro FTYPE, to test types other than off_t.
(_AC_SYS_LARGEFILE_TEST_CODE): Test ino_t for
-D_FILE_OFFSETBITS=64 too, if no flags are needed for off_t.
Needed for GNU/Linux on alpha and s390x.
* lib/autoconf/specific.m4 (AC_SYS_YEAR2038_RECOMMENDED):
Do not recommend gcc -m64, as that likely will not work.
Problem reported by Bruno Haible in:
https://lists.gnu.org/r/bug-gnulib/2023-05/msg00060.html
Instead, mention that 32-bit time_t is not recommended
for this package, before telling the builder how
to configure with 32-bit time_t anyway.
Problem reported by Matt Turner in:
https://lists.gnu.org/r/bug-autoconf/2023-05/msg00005.html
* lib/autoconf/functions.m4 (AC_FUNC_MMAP): Go back to getting the
page size, since the zero-fill test needs this after all.
However, prefer sysconf (_SC_PAGESIZE) or sysconf (_SC_PAGE_SIZE)
to getpagesize (), and use ‘long’ not ‘int’ to store the page size.
Also, declare getpagesize if it is used as a function.
* lib/autoconf/functions.m4 (AC_FUNC_MMAP): Don’t call
getpagesize, as it’s tricky to configure, modern POSIX doesn’t
require it, and calling it without including <unistd.h> provokes a
compile-time error on modern systems. Instead, rework the test
to not need getpagesize. Add a FIXME comment for unnecessary
tests; I don't want to remove them now as we're too close to
a release. Remove long-obsolete comment about GNU grep.
New macro AC_SYS_YEAR2038_RECOMMENDED replaces new macro
AC_SYS_YEAR2038_REQUIRED, and gives the builder an out of
specifying --disable-year2038. Remove new macro
AC_SYS_LARGEFILE_REQUIRED, which was added only for symmetry and
does not seem to have a great need.
* NEWS, doc/autoconf.texi: Document this.
* lib/autoconf/specific.m4:
Be more specific about mid-January 2038 than just Jan 2038.
(_AC_SYS_YEAR2038_PROBE): Ignore IF-NOT-DETECTED arg.
If support is not detected, merely set ac_have_year2038=no instead
of erroring out. All callers changed.
(_AC_SYS_YEAR2038_OPT_IN): Remove. All callers removed.
(AC_SYS_YEAR2038): Simplify by requiring AC_SYS_LARGEFILE
and then testing the result.
(AC_SYS_YEAR2038_REQUIRED, AC_SYS_LARGEFILE_REQUIRED): Remove.
(AC_SYS_YEAR2038_RECOMMENDED): New macro.
(_AC_SYS_LARGEFILE_PROBE): If support is not detected, merely set
ac_have_largefile=no instead of erroring out. All callers changed.
Take on the burden of invoking year2038 probe as needed.
(AC_SYS_LARGEFILE): Simplify.
* lib/autoconf/specific.m4 (_AC_SYS_LARGEFILE_PROBE): Distinguish the results
"support not detected" and "supported through gnulib". If the result is
"supported through gnulib", don't fail.
BusyBox 1.35.0 tr, which is shipped with Alpine Linux 3.17, does not support
the POSIX [x*n] syntax.
* doc/autoconf.texi (Limitations of Usual Tools): Document limitation of 'tr'
from BusyBox.
* lib/autoconf/specific.m4 (_AC_SYS_YEAR2038_PROBE)
(_AC_SYS_YEAR2038_ENABLE, _AC_SYS_YEAR2038_OPT_IN)
(AC_SYS_YEAR2038, _AC_SYS_LARGEFILE_PROBE)
(_AC_SYS_LARGEFILE_ENABLE): Do not use enable_largefile to record
whether largefile is required, as ‘./configure
--disable-largefile’ sets enable_largefile=no even if largefile is
required and this disables largefile. Instead, use a separate
shell variable ac_largefile_required and test it as well.
Similarly for enable_year2038.
(_AC_SYS_LARGEFILE_ENABLE): Omit --disable-largefile help string
if year2038 is required, since largefile is a prereq for year2038.
Problem reported by Frederic Berat in:
https://lists.gnu.org/r/bug-rcs/2023-04/msg00001.html
* lib/autoconf/specific.m4 (_AC_SYS_LARGEFILE_TEST_INCLUDES):
Bring back for compatibility with post-2.71, pre-2.73 Gnulib.
Most of the remaining uses were converted to AC_COMPILE_IFELSE.
The use in AC_FUNC_LOADAVG becomes an AC_PREPROC_IFELSE because
we can’t be sure getloadavg.c can be _compiled_ at this point in
the build. The use in AC_C_VARARRAYS could be either _PREPROC_ or
_COMPILE_; we use _COMPILE_ because, _PREPROC_ is never used, then
we don’t have to do the “checking how to run the C preprocessor” test.
* lib/autoconf/c.m4 (AC_C_VARARRAYS): Use AC_COMPILE_IFELSE instead of
AC_EGREP_CPP.
* lib/autoconf/headers.m4 (_AC_HEADER_TIOCGWINSZ_IN_TERMIOS_H)
(_AC_HEADER_TIOCGWINSZ_IN_SYS_IOCTL_H): Likewise.
* lib/autoconf/functions.m4 (AC_FUNC_GETLOADAVG): Use AC_PREPROC_IFELSE
instead of AC_EGREP_CPP.
Another of the last few uses of AC_EGREP_CPP is to check for the ISO C
“stringification” operator. As this is a feature of C89, let’s fold
the test into the extensive C89 tests we already have, and make
AC_C_STRINGIZE just lean on AC_PROG_CC, in the same way
AC_C_PROTOTYPES does.
* lib/autoconf/c.m4 (_AC_C_C89_TEST_GLOBALS): Add test of preprocessor
stringification and token pasting.
(AC_C_STRINGIZE): Just check ac_prog_cc_stdc.
AC_TYPE_GETGROUPS is the last remaining use of AC_EGREP_HEADER in
stock Autoconf macros. It uses it only when cross compiling, as a
fallback from an AC_RUN_IFELSE check, testing for a bug in system
headers from the late 1980s or early 1990s, where gid_t *existed*
but the second argument to getgroups needed to be an array of int,
and this didn’t cause a compile error (i.e. the system headers
declare getgroups with no prototype or an incorrect prototype).
AC_FUNC_GETGROUPS also uses AC_RUN_IFELSE to test for obscure
problems specific to long-obsolete Unixes.
The downsides of AC_RUN_IFELSE and AC_EGREP_HEADER seem more severe
than the chances of someone compiling a current-generation program,
that uses getgroups, on an OS old enough to have one of the really
nasty bugs. Accordingly, this patch changes AC_FUNC_GETGROUPS to use
a host_os-based *blacklist* both in native and cross compilation.
This is limited to the two host_os values for which either our old
code, or Gnulib, documented a serious bug: ultrix* and nextstep*.
Currently it does not try to pin down the exact version ranges subject
to the bugs — that would require research by someone with access to
the full history of these OSes.
An incorrect guess by this blacklist can be overridden by setting
ac_cv_func_getgroups_works in config.site. AC_TYPE_GETGROUPS, for its
part, now does a series of regular old AC_COMPILE_IFELSE checks to
probe the prototype of getgroups, and considers that good enough.
While I was in there I noticed that AC_FUNC_GETGROUPS does not
AC_SUBST a documented output variable, and that the name of this
variable is misspelled in the manual.
* lib/autoconf/functions.m4 (AC_FUNC_GETGROUPS): Use AC_SEARCH_LIBS
to probe for getgroups. Use an AC_CANONICAL_HOST-based blacklist
for bug detection, not AC_RUN_IFELSE. AC_SUBST the GETGROUPS_LIB
output variable.
* lib/autoconf/types.m4 (AC_TYPE_GETGROUPS): Check only the prototype
of getgroups, using AC_COMPILE_IFELSE; do not use either AC_RUN_IFELSE
or AC_EGREP_HEADER.
* doc/autoconf.texi: Update to match. Correct misspelling of
GETGROUPS_LIB.
* tests.local.at (_AT_CHECK_ENV): Allow GETGROUPS_LIB output variable.
This macro is one of the last remaining internal uses of AC_EGREP_CPP.
It has only ever done anything useful with GCC, and GCC dropped
support for ‘traditional’ compilation in version 3.3 (released 2003)
so I do not think it is worth trying to preserve.
* lib/autoconf/c.m4 (AC_PROG_GCC_TRADITIONAL): Make into a
compatibility alias for AC_PROG_CC, similar to AC_PROG_CC_STDC.
* lib/autoconf/general.m4 (AC_EGREP_CPP): Remove stale comment.
* doc/autoconf.texi, NEWS: Document this change.
* tests/mktests.pl: Exclude AC_PROG_GCC_TRADITIONAL from
autoupdate tests.
AC_TYPE_UID_T uses AC_EGREP_HEADER to search sys/types.h for
occurrences of the string ‘uid_t’ and, if found, assumes both
uid_t and gid_t are available. This would be better done using
a pair of AC_CHECK_TYPE operations.
I also converted two uses of old-style AC_CHECK_TYPE, immediately
below, to new-style. (There are probably other old-style uses in
this file, I only did the ones I happened to see.)
* lib/autoconf/types.m4 (AC_TYPE_UID_T): Check for uid_t and gid_t,
separately, using AC_CHECK_TYPE, instead of grepping sys/types.h.
(AC_TYPE_SIZE_T, AC_TYPE_SSIZE_T): Use new-style AC_CHECK_TYPE.
AC_XENIX_DIR is an obsolete macro, defined as AC_HEADER_DIRENT plus
code to make absolutely sure configure scripts that depended on a
shell variable internal to the original (2.13 era) definition of
AC_XENIX_DIR are not broken by autoupdate. (That variable had the
temptingly public-sounding name “XENIX.”) This compatibility code
uses AC_EGREP_CPP, which is itself discouraged for use in new
configure scripts.
(N.B. codesearch.debian.net does not find any uses whatsoever of
this macro, nor any code in an .ac or .m4 file that depends on the
XENIX variable.)
Change the compatibility code to use AC_CANONICAL_HOST instead,
and clarify which pieces of the code inserted by autoupdate are
probably still necessary.
* lib/autoconf/specific.m4 (AC_XENIX_DIR): Set XENIX variable
based on value of host_os. Clarify what manual cleanup is
recommended after autoupdate replaces this obsolete macro.
Problem reported bh Khem Raj for mcpp 2.7.2 (2008) in:
https://lists.gnu.org/r/autoconf/2023-04/msg00001.html
* lib/autoconf/programs.m4 (_AC_PATH_PROG_FLAVOR_GNU):
Add two ‘@%:@(’s to cater to underquoted callers.
* lib/m4sugar/m4sh.m4 (_AS_PATH_WALK): Use quadrigraph instead of
‘#’, for underquoted callers.
Problem reported by Bogdan via Jacob Bachmeyer in:
https://lists.gnu.org/r/autoconf/2023-04/msg00002.html
* bin/autom4te.in: If a file timestamp equals a dependency’s
timestamp, consider the file to be out of date. Although this may
result in extra work, it fixes some rare timing bugs.
Commit 61901a1a14 dated 2022-07-10
bumped the Perl requirement to 5.10 or later, because
commit 3a9802d601 dated 2021-08-31
added code using Time::HiRes’s ‘stat’ function, a feature
added in Perl 5.8.9+ or Perl 5.10+, and it was hard
to find Perl 5.8.9 hosts to test with. Also, requiring Perl 5.10
meant that we could then use operators like Digest::SHA, the // and
//= operators, the regexp \K escape, and ‘state’ variables.
However, that Time::HiRes code, which was taken from Automake, has
recently been made optional by Automake, and it now works again with
Perl 5.6. And Autoconf is not yet using any other post-5.6 feature,
except when developers run help-extract.pl (something Autoconf users
do not use). So relax the Autoconf user requirement back to 5.6 as it
was in Autoconf 2.71; although Autoconf developers will need 5.10 or
better, Autoconf users can get by with 5.6.
I ran into this problem when testing the Autoconf release candidate on
Solaris 10, which has Perl 5.8.4. Oracle says Solaris 10’s
end-of-life is January 2024, so it’s still (barely) a viable porting
target. Of course with Solaris 10 one must install a recent-enough
GNU m4, but adding a requirement to also install a recent-enough Perl
is a new barrier, and if it’s not needed then it might be better to
wait until it is needed (or until 2024 arrives).
* NEWS: Update news item about Perl 5.6 vs 5.10.
* README-hacking: Bump Perl recommendation to 5.10.
* build-aux/fetch.pl: Do not munge imported code to require 5.10.
* tests/autotest.at (parallel autotest and signal handling): This
test would consistently fail due to an exit status of 0. That was
considered failure because the test required a SIGHUP-indicating
exit status. However, an status of 0 is perfectly fine, too, so
accept that.
Preparing to make a pre-release snapshot, update these:
* build-aux/announce-gen: Update from gnulib
* build-aux/gnupload: Likewise.
* build-aux/texinfo.tex: Update from texinfo.
‘guix shell’ is a utility for constructing isolated environments
for development; I’ve put together a “manifest” stating the
build and test requirements for autoconf, starting from a git
checkout.
If we use a relative path for ACBOOTDIR, Automake can’t tell the
difference between Autoconf’s configure script’s aclocal.m4
inclusions (…/m4/*.m4) and the guts of Autoconf
itself (…/lib/autoconf/*.m4) so it puts both of them into
$(am__aclocal_m4_deps). This would be harmless, except that the
guts-of-Autoconf files are named *relative to $ACBOOTDIR*, which
means Make can’t find them later. And this is why a build from a
clean git checkout always starts by regenerating aclocal.m4 and
configure again.
Using an absolute path for ACBOOTDIR gives automake enough of a clue
what’s going on (see the heuristic circa 5500 of current automake.in,
commented “Some modified versions of autoconf don’t use frozen files…”)
for it to produce the same value for $(am__aclocal_m4) that it would
if we were running an installed Autoconf.
It is not clear to me why, but the “parallel autotest and signal
handling” test malfunctions if the active shell is Guix bash. I don’t
think it’s worth investigating in detail, considering I intend to
reimplement parallel autotest using the same technique that Automake’s
parallel test driver uses, i.e. make -j, which should eliminate this
entire class of problems.
* cfg.mk (local-checks-to-skip): Add sc_unportable_grep_q, which has
too many false positives to bother with; for instance, it triggers
on autoconf.texi’s discussion of why grep -q is unportable, and on
the code in maint.mk that implements the check!
(old_NEWS_hash): Update for commit b751bf4949,
which fixed spelling errors in old NEWS.
* doc/autoconf.texi: Remove a doubled word.
* lib/autoconf/programs.m4: Remove a space immediately before a tab.
* lib/m4sugar/m4sh.m4 (_AS_IF): Rephrase documentation to avoid saying
“if IF-FALSE” which triggers the prohibit_doubled_word check.
