The list of macros documented as being defined by
AC_USE_SYSTEM_EXTENSIONS had gotten out of sync with the actual list.
Update it thoroughly.
Also, I introduced an error into the commentary when I merged Julien
ÉLIE’s patch to define _NETBSD_SOURCE and _OPENBSD_SOURCE in
AC_USE_SYSTEM_EXTENSIONS. _OPENBSD_SOURCE does something on NetBSD
and *doesn’t* do anything on OpenBSD. This is corrected.
Clean up the code in AC_USE_SYSTEM_EXTENSIONS a bit while I’m in
there; we now had a redundant definition of _NETBSD_SOURCE (one
unconditional and one conditional on minix/config.h existing).
Reorganize the macro to make it easier to catch problems like this in
the future.
* lib/autoconf/specific.m4 (AC_USE_SYSTEM_EXTENSIONS): Reorganize;
remove redundant AC_DEFINE of _NETBSD_SOURCE; add some missing
AC_BEFOREs; use _AC_CHECK_HEADER_ONCE for header checks;
revise all commentary.
* doc/autoconf.texi (AC_USE_SYSTEM_EXTENSIONS): Update.
Makes the documentation of AC_PROG_CC consistent with the
documentation of AC_PROG_CXX. Also removes a bunch of redundant text
from c.m4 and adds lists of the headers that *can* be used in the
conformance tests, so future hackers don’t have to look them up.
* doc/autoconf.texi (AC_PROG_CC): Make description consistent with
description of AC_PROG_CXX.
* lib/autoconf/c.m4: Clean up some outdated or repetitive commentary
and add lists of the freestanding headers above the code that needs
to avoid using non-freestanding headers.
This makes the C conformance tests more consistent with the C++
conformance tests, and should also speed up cycling through the
possible options to turn on C99/C11.
Tested with gcc, clang, SunPRO C, and AIX xlc.
* lib/autoconf/c.m4 (_AC_C_C89_TEST_GLOBALS): Add preprocessor test
for __STDC__ being defined (to any value).
(_AC_C_C99_TEST_GLOBALS, _AC_C_C11_TEST_GLOBALS): Add preprocessor
test of the value of __STDC_VERSION__.
In testing on Darwin (OSX), sometimes warnings reported from M4 code
reach autom4te with no stack trace at all, causing the perl script to
crash with a “use of uninitialized value” error. The root cause of
the problem is not clear to me, but the script certainly shouldn’t
crash.
Problem found by Jannick <thirdedition@gmx.net>.
* bin/autom4te.in: When processing warnings, make sure $stacktrace is
defined.
* tests/atlocal.in: Also set AWK to value detected by configure.
Alphabetize list of shell variables set by config.status.
* tests/local.mk: Add a rule to recreate tests/atconfig when
config.status changes.
The tests for the level of the C and C++ standard supported by their
respective compilers should also avoid using any headers that are not
guaranteed to be available in the respective freestanding environment.
Unlike the previous change, the only user-visible consequence of this
one should be that C11/C99/C89/C++11/C++98 *compiler* support is now
correctly detected when the compilation target is a freestanding
environment.
This patch also refactors how we “emit [the text of the C/C++
standard-conformance test programs] only once per [configure script],
into shell variables which can then be referenced repeatedly,” from
c3853873, because editing them just a little made the M4 quotation
break. Clearly too fragile.
I believe this completes the fix for bug #110393.
* lib/autoconf/c.m4 (_AC_PROG_CC_C89, _AC_PROG_CC_C99, _AC_PROG_CC_C11)
_AC_C_C99_TEST_HEADER, _AC_C_C99_TEST_BODY): Move all test program
fragments into new macros that can be AC_REQUIREd individually:
_AC_C_C89_TEST_GLOBALS, _AC_C_C89_TEST_MAIN, _AC_C_C89_TEST_PROGRAM,
_AC_C_C99_TEST_GLOBALS, _AC_C_C99_TEST_MAIN, _AC_C_C99_TEST_PROGRAM,
_AC_C_C11_TEST_GLOBALS, _AC_C_C11_TEST_MAIN, _AC_C_C11_TEST_PROGRAM.
Each emits test code at most once, into a shell variable in the
INIT_PREPARE diversion.
Revise each test program to use only library features of the
respective standard’s freestanding environment.
(_AC_C_STD_TRY): Take the *name* of the shell variable holding the
complete test program as an argument, not the code itself. All
callers adjusted to match.
(_AC_PROG_CXX_CXX98, _AC_PROG_CXX_CXX11, _AC_CXX_STD_TRY)
(_AC_CXX_CXX98_TEST_HEADER, _AC_CXX_CXX98_TEST_BODY)
(_AC_CXX_CXX11_TEST_HEADER, _AC_CXX_CXX11_TEST_BODY): Similarly.
New macros are:
_AC_CXX_CXX98_TEST_GLOBALS, _AC_CXX_CXX98_TEST_MAIN,
_AC_CXX_CXX98_TEST_PROGRAM,
_AC_CXX_CXX11_TEST_GLOBALS, _AC_CXX_CXX11_TEST_MAIN,
_AC_CXX_CXX11_TEST_PROGRAM.
Since 1993, Autoconf has been assuming that it is safe to include any
of the headers defined by ISO C90 without checking for them; this is
inaccurate, since only a subset are necessarily available in a
C90 *freestanding* environment.
It is OK to assume the presence of a header in a macro that checks
specifically for something declared by that header (if the header is
not present, we will think the specific declaration is unavailable,
which is probably accurate for modern embedded environments). It is
also OK to continue recommending that user code use these headers
unconditionally—anyone working with a freestanding environment knows
it. But it is not OK for very generic code within Autoconf itself,
such as AC_INCLUDES_DEFAULT, to make this assumption.
Note that the set of headers that are not always available includes
stdio.h, which we have been assuming can be included unconditionally
for even longer.
In AC_INCLUDES_DEFAULT, revert to checking for string.h and stdlib.h
before including them. Also revert to defining STDC_HEADERS only when
string.h and stdlib.h are available (but do not check for float.h and
stdarg.h, as these are part of the freestanding set). Add a new check
for stdio.h. Sort the inclusion list by standard (C90 freestanding;
C90 hosted; C99; POSIX) and alphabetically within each group. Revise
all the documentation and update the testsuite.
This partially reverts commit 86c213d0e3
and is a partial fix for bug #110393.
* lib/autoconf/headers.m4 (AC_CHECK_INCLUDES_DEFAULT): Check for
stdio.h, stdlib.h, and string.h before including them. Define
STDC_HEADERS only when string.h and stdlib.h are both available.
Organize includes list by standard, then alphabetically.
* doc/autoconf.texi, NEWS: Update to match.
* tests/local.at (AT_CHECK_DEFINES): Make regexes more specific.
Also expect a definition of HAVE_STDIO_H.
* tests/c.at, tests/semantics.at, tests/tools.at: Use <float.h>,
not <stdio.h>, as a header that we expect always to exist.
Add HAVE_STDIO_H to various lists of macros that are expected to
appear in config.h.
Revert commit 18c140b50b, restoring
AC_PROG_CC to being defined as an ordinary AC_DEFUN. This broke
third-party macros (e.g. the Autoconf Macro Archive’s
AX_PROG_CC_FOR_BUILD) that intentionally invoked AC_PROG_CC a second
time with its guts redefined via a whole bunch of ‘pushdef’s. I don’t
think we want to support this long-term, but needing access to a
build-native compiler in cross-compilation is common enough that we
should have *some* supported way to do it, and it may as well be
AX_PROG_CC_FOR_BUILD until we come up with something better.
If we go back to AC_DEFUN_ONCE for AC_PROG_CC in the future, we should
do it consistently for all the “find me a compiler” macros -- it
was *only* done for AC_PROG_CC in 18c140b5.
The rationale for AC_DEFUN_ONCE seems to have been to reduce the size
of the generated configure script. The bulk of the size accountable to
AC_PROG_CC is the test programs for figuring out which version of the
C standard is available, so I tweaked _AC_C_STD_TRY (and _AC_CXX_STD_TRY)
to emit that text only once per program, into shell variables which
can then be referenced repeatedly.
Fixes bug #110350.
* NEWS, doc/autoconf.texi: Revert documentation changes associated
with AC_PROG_CC being a one-shot macro.
* lib/autoconf/c.m4 (AC_PROG_CC): Revert to defining with AC_DEFUN.
(_AC_C_STD_TRY, _AC_CXX_STD_TRY): Emit the test program only once,
even if invoked multiple times with the same arguments.
* tests/foreign.at (AX_PROG_CC_FOR_BUILD, AX_PROG_CXX_FOR_BUILD):
New tests.
Currently, there isn’t any documented way for an Autotest testsuite to
add custom code to be run either right before the main driver loop, or
at the point of each AT_SETUP. For instance, there’s no good place to
put environment variable sanitization that should apply to the entire
testsuite (but isn’t universally relevant), or shell function
definitions to be used by custom test macros.
Autoconf’s test suite is poking shell functions directly into the
PREPARE_TESTS diversion, and doing environment variable sanitization
in each individual test. Both of these are obviously undesirable.
This patch adds three new AT_* macros that can be used to do these
things in an officially-supported way: AT_PREPARE_TESTS adds code to
be run right before the main driver loop, AT_PREPARE_EACH_TEST adds
code to be run at the beginning of each test, and AT_TEST_HELPER_FN
defines a shell function that will be available to each test. In
Autoconf’s test suite, I use AT_PREPARE_TESTS to factor out
environment variable sanitization that *ought* to apply across the
board, and AT_TEST_HELPER_FN for the helper function used by
AT_CHECK_ENV.
(This fixes the testsuite bug reported by Jannick at
https://lists.gnu.org/archive/html/autoconf/2020-10/msg00052.html :
CONFIG_SITE in the parent environment will no longer be visible to tests.)
It would be nice to give an example of when AT_PREPARE_EACH_TEST is
useful, in the documentation, but I didn’t find one in the autoconf
test suite.
* lib/autotest/general.m4 (AT_PREPARE_TESTS, AT_PREPARE_EACH_TEST)
(AT_TEST_HELPER_FN): New macros.
(AT_INIT, AT_TESTED): Emit the code to report tested programs only
if it’s needed, and make sure it’s after any code added by
AT_PREPARE_TESTS.
* tests/local.at: Add AT_PREPARE_TESTS block that ensures
$MAKE is set sensibly and $MAKEFLAGS and $CONFIG_SITE are unset.
Use AT_TEST_HELPER_FN for the helper function needed by AT_CHECK_ENV.
(AT_CHECK_MAKE): No need to sanitize $MAKE or $MAKEFLAGS here.
* tests/base.at, tests/compile.at, tests/m4sh.at, tests/torture.at:
No need to unset or neutralize $CONFIG_SITE in individual tests.
* tests/autotest.at: Add tests for new macros.
* doc/autoconf.texi, NEWS: Document new macros.
* doc/autoconf.texi (Preset Output Variables): Clarify that options -m32 or -m64
must go into CC, not CFLAGS and not CPPFLAGS either. This is needed because on
bi-arch platforms, config.guess runs $CC without $CFLAGS nor $CPPFLAGS.
The new feature of autoreconf --install installing config.guess,
config.sub, and install-sh itself didn’t implement --force mode
correctly; existing files would not be overwritten.
The fix has two parts. If --force is in effect, we try to install
everything that we can from the needed-auxfiles list *before* checking
which of them already exist. But also, we don’t overwrite existing
files unconditionally, even with --force; we only do so if the file we
can install has a newer “timestamp” line than the copy at the
destination. This is because Automake can also install all of the
files we can install. Suppose someone is using autoconf 2.70 with a
newly released automake 1.17; automake 1.17 will presumably ship with
a newer config.guess than autoconf 2.70 did; that version should win.
Fixes bug #110371.
* bin/autoreconf.in (extract_time_stamp, our_aux_file_is_newer): New functions.
(install_aux_file): If the destination exists, and our copy is not
newer, do not overwrite it.
(autoreconf_current_directory): When $force is true, attempt to
install all needed aux files *before* scanning for missing aux files.
(throughout): Remove extra \n from various error messages.
* tests/torture.at (Missing auxiliary files (--force)): New test.
Many of the reported regressions in Autoconf 2.70 betas went unnoticed
for years because Autoconf’s bundled test suite didn’t test most of
the macros with a C++ compiler and/or in cross compilation mode.
There’s a special makefile target ‘maintainer-check-c++’ that runs all
the tests with CC=g++, but that doesn’t catch the regressions either,
because it doesn’t compare the configure results with what you’d have
gotten with a C compiler. Also, C and C++ have diverged to the point
where setting CC to a C++ compiler doesn’t work reliably anymore.
This patch overhauls AT_CHECK_MACRO to test each macro four times:
(C compiler, C++ compiler) x (native mode, cross-compilation mode).
All four tests are expected to produce the same config.cache and
config.h, except for certain predictable differences due to running
AC_PROG_CXX instead of AC_PROG_CC, and a short list of known,
acceptable differences, maintained in mktests.pl.
There are two classes of known, acceptable differences. Macros that
use AC_RUN_IFELSE aren’t tested in cross-compilation mode at all,
because they may crash the script (this is temporary and will be
revisited after 2.70). Macros that correctly detect a difference
between C and C++ (e.g. AC_HEADER_STDBOOL will notice that C++ doesn’t
have the _Bool type) are annotated with the specific cache variable
and #define that varies.
mktests.pl now also has the capability to provide values for the
MACRO-USE, ADDITIONAL-COMMANDS, and AUTOCONF-FLAGS arguments to
AT_CHECK_(AU_)MACRO, on a per-macro basis, but that’s not used in this
patch.
Some of the manual uses of AT_CHECK_MACRO do not need to test C++
and/or cross compilation; for them, there is a new test helper,
AT_CHECK_CONFIGURE_AC. Another new helper, AT_PRESERVE_CONFIG_STATUS,
is used extensively in AT_CHECK_(AU_)MACRO but may be also useful in
manual tests that need to do multiple configure runs.
This change supersedes AT_CHECK_MACRO_CROSS and
‘make maintainer-check-c++’, which are removed.
In my testing, setting CC to a C++ compiler no longer works at all,
for reasons that are impractical to fix (e.g. C++ compilers choke on
the test for C2011 features) so I have added a note to NEWS saying
that this is not supported anymore.
* tests/local.at (AT_CHECK_MACRO): Default behavior is now to test
the macro in both native and cross-compilation mode, and expect the
results to be identical. If the macro transitively required
AC_PROG_CC, and a C++ compiler is available, then test it twice
more with AC_LANG([C++]) in effect, and again expect the results to
be identical. New fifth argument TEST-PARAMETERS can modify this
behavior.
(_AT_FILTER_CXX_CV_VARIES, _AT_FILTER_CXX_DEFINE_VARIES): New,
subroutines of AT_CHECK_MACRO.
(AT_CHECK_MACRO_CROSS): Remove, subsumed by new AT_CHECK_MACRO
behavior.
(AT_CHECK_AU_MACRO): Forward to AT_CHECK_MACRO for the basic test;
then do the same autoupdate test as before, as a separate test group.
(at_check_env): Also ignore OPENMP_CXXFLAGS.
(AT_CONFIG_CMP): Add third argument EXTRA-VARIANCE that specifies
additional variables that are expected to vary in a particular test.
(_AT_CONFIG_CMP_PRUNE): New, subroutine of AT_CONFIG_CMP.
(AT_DEFINES_CMP): New helper macro that compares config.h headers,
with the ability to ignore variation in specific defines.
(_AT_DEFINES_CMP_PRUNE): New, subroutine of AT_DEFINES_CMP.
(AT_PRESERVE_CONFIG_STATUS): New helper that makes copies of
config.h, config.log, config.status, and state-env.after under
names that won’t be clobbered by a subsequent run of configure.
(AT_CHECK_CONFIGURE_AC): New helper that defines a complete test
group consisting of a single invocation of _AT_CHECK_AC_MACRO;
effectively what AT_CHECK_MACRO used to be.
(_AT_CHECK_AC_MACRO): Correct documentation comment; the PRE-TESTS
argument has always been optional.
* tests/mktests.pl (test_parameters): New global data object giving
extra arguments to pass to AT_CHECK_MACRO/AT_CHECK_AU_MACRO on a
per-macro basis.
(emit_test): New function that handles emitting calls to
AT_CHECK_(AU_)MACRO with the desired arguments.
(scan_m4_files): Use emit_test.
(au_exclude_list): Add AC_HAVE_LIBRARY, AC_COMPILE_CHECK,
AC_TRY_CPP, AC_TRY_COMPILE, AC_TRY_LINK, and AC_TRY_RUN.
* tests/semantics.at (AC_CHECK_LIB, AC_SEARCH_LIBS): Rewrite test
using symbols from zlib instead of libm, to get consistent behavior
from C and C++.
(AC_SEARCH_LIBS (none needed)): Revise to clarify what is being tested.
(AC_CHECK_DECLS): Use _AC_LANG_ABBREV when inspecting cache variables.
(AC_CHECK_ALIGNOF, AC_CHECK_ALIGNOF struct)
(AC_CHECK_SIZEOF, AC_CHECK_SIZEOF struct)
No need for AT_CHECK_MACRO_CROSS.
(AC_CHECK_FILES): Switch to AT_CHECK_CONFIGURE_AC.
(AC_SYS_RESTARTABLE_SYSCALLS, AC_FUNC_WAIT3): Do not test in cross
compilation mode.
(AC_TRY_CPP, AC_TRY_COMPILE, AC_TRY_LINK, AC_TRY_RUN)
(AC_COMPILE_CHECK, AC_HAVE_LIBRARY): New manual AT_CHECK_AU_MACRO tests.
* tests/c.at (Extensions, C keywords, AC_PROG_CPP requires AC_PROG_CC)
(AC_NO_EXECUTABLES (working linker), AC_NO_EXECUTABLES (broken linker)):
Switch to AT_CHECK_CONFIGURE_AC. Also convert case statements to AS_CASE.
(Broken/missing compilers): Pass CC=no-such-compiler on the command
line instead of hardwiring it in the configure script.
* tests/local.mk (maintainer-check-c++): Remove target.
(maintainer-check): Run the ordinary ‘make check’ as well as
‘make maintainer-check-posix’.
This is probably marginally faster since it doesn’t have to read the
.m4 files twice or muck around with temporary files and subprocesses,
but the actual point of this rewrite is that it will make it easier to
provide additional arguments to AT_CHECK_MACRO and AT_CHECK_AU_MACRO
on a per-macro basis; this capability will be added and used in the
next patch.
In *this* patch, the ac*.at files are not functionally changed at all.
(The comments come out slightly differently, though.)
* tests/mktests.sh: Delete and...
* tests/mktests.pl: ...rewrite in Perl.
* tests/local.mk: Update to match.
Erlang is similar to Java in that it doesn’t compile to standalone
machine code; the output of ‘erlc’ is byte-code files that are then
interpreted by ‘erl’. We handle this poorly in a whole bunch of ways,
particularly when cross-compiling. This patch fixes up the more
serious problems:
- AC_COMPILE_IFELSE now actually works when AC_LANG([Erlang]) is in
effect.
- ‘conftest.beam’ is now deleted in several more places where it
could be created.
- The various AC_ERLANG_* macros that interrogate the runtime
environment do so by invoking ‘$ERL’ directly, rather than using
AC_RUN_IFELSE, and thus do not crash the configure script when
we think we’re cross-compiling. (It is not clear to me whether
they get the correct answer when cross-compiling, but this should
still be strictly an improvement.)
- The Erlang-related tests have been streamlined.
Further improvements are definitely possible, but we’d have to teach
the infrastructure to make $ac_objext language-specific first, which
seems like too big of a change for 2.70.
(This patch is all fallout from a logically unrelated testsuite change
which is coming up next. Gotta love the fundamental interconnectedness
of things.)
* lib/autoconf/general.m4 (_AC_COMPILE_IFELSE_BODY)
(_AC_LINK_IFELSE_BODY): Delete conftest.beam as well as conftest.$ac_objext.
* lib/autoconf/erlang.m4 (AC_ERLANG_PATH_ERLC, AC_ERLANG_PATH_ERL):
Don’t repeat work done by AC_PATH_TOOL.
(Erlang $ac_compile): Fake an .o file so AC_TRY_COMPILE will be happy.
(AC_LANG_COMPILER(Erlang)): AC_REQUIRE AC_ERLANG_NEED_ERLC, not
AC_ERLANG_PATH_ERLC. Also AC_REQUIRE AC_ERLANG_NEED_ERL so
AC_RUN_IFELSE works reliably.
(AC_ERLANG_CHECK_LIB, AC_ERLANG_SUBST_ROOT_DIR)
(AC_ERLANG_SUBST_LIB_DIR, AC_ERLANG_SUBST_ERTS_VER):
Use $ERL -eval, not AC_RUN_IFELSE.
No need to AC_REQUIRE AC_ERLANG_NEED_ERLC.
* tests/erlang.at: Don’t test anything here that’s tested adequately
by acerlang.at; document which macros those are expected to be.
Remove unnecessary AC_ERLANG_PATH_ERL/ERLC invocations throughout.
(AT_CHECK_MACRO([Erlang])): Rename test to ‘Erlang basic compilation’;
expect both AC_COMPILE_IFELSE and AC_RUN_IFELSE to work;
handle cross compilation mode properly.
* tests/mktests.sh: Exclude from acerlang.at all macros completely
covered by erlang.at.
M4-redefining ‘printf’ as ‘echo’ brings back all the variations in
‘echo’ behavior that we were trying to get away from by switching to
‘printf’ in the first place. This caused a spurious failure on AIX.
* tests/m4sh.at (Redefining AS_ECHO internals): Redefine ‘printf’ as
a shell function with fully predictable output, not as ‘echo’.
All tests that run autoreconf need to defend against the possibility
of aclocal not existing and/or barfing on bugs in third-party m4 files
installed on the build system. Two of the “Missing auxiliary files”
tests were missing this defensive code.
* tests/torture.at (Missing auxiliary files (install-sh))
(Missing auxiliary files (foreign)): Prevent autoreconf from
running aclocal.
Despite what the documentation says, ‘new File::Temp’ does not work
reliably in perl 5.6.x. Rather than figure out exactly what is wrong
with it, let’s just stick to ‘tempfile’.
* bin/autom4te.in (handle_output): Use tempfile function instead of
object-oriented File::Temp interface.
* bin/autoreconf.in (install_aux_file): Likewise.
AC_CHECK_HEADER_STDBOOL is documented to make two checks: whether the
C99 header <stdbool.h> is available and fulfills its
specification (i.e. including it makes the type ‘bool’ and the
constants ‘true’ and ‘false’ available), and, independently, whether
the type ‘_Bool’ is available.
In C++, the type ‘_Bool’ is usually _not_ available, but <stdbool.h>
is still supposed to be include-able and the type ‘bool’ and the
constants ‘true’ and ‘false’ are still supposed to be available
(unconditionally). However, the test for <stdbool.h> fulfilling its
specification freely used _Bool, and would therefore fail spuriously.
Correct this by checking for _Bool first, and then refactoring the
test program for <stdbool.h> so that it does all its tests using bool,
then repeats them with _Bool only when available.
* lib/autoconf/headers.m4 (AC_CHECK_HEADER_STDBOOL): Do the test for
_Bool before the test for stdbool.h. Test semantics of bool
unconditionally; test _Bool only when HAVE__BOOL is defined.
AC_FUNC_SETPGRP determines whether you have the historic BSD setpgrp,
which takes two arguments and returns int, or the historic POSIX
setpgrp, which takes no arguments and returns int. Solaris has yet a
third variant, which takes no arguments and returns a pid_t (the new
process group ID). This difference causes AC_FUNC_SETPGRP’s test
program to fail to compile under AC_LANG([C++]), which in turn causes
the macro to report that setpgrp does take arguments, which is wrong.
It is not worth adding a new result #define for this variant,
since *all* forms of setpgrp are deprecated in favor of setpgid, which
is old enough that it can be used unconditionally. However, it is
worth documenting that this variant exists, and fixing AC_FUNC_SETPGRP
to produce the right value for its existing result #define on Solaris
with C++.
* lib/autoconf/functions.m4 (AC_FUNC_SETPGRP): Redesign test program to
not depend on the return type of setpgrp.
* doc/autoconf.texi (AC_FUNC_SETPGRP): Mention that the macro does not
check for the Solaris variant of setpgrp that returns pid_t. Change
programming advice to recommend use of setpgid.
On systems where plain ‘char’ is unsigned (e.g. AIX) we would define
__CHAR_UNSIGNED__ only when $GCC was not true at configure time.
If AC_LANG([C++]) has been in effect since the beginning of the
script (so AC_PROG_CC was never invoked), $GCC will be false
regardless; this causes an inconsistency between the C and C++
behaviors, even when both compilers are GNU.
The point of checking $GCC here is that GCC has command line options
to override the signedness of plain ‘char’, and it predefines
__CHAR_UNSIGNED__ to indicate what the signedness actually is.
We don’t want config.h to override that. However, there is already
a special autoheader template for __CHAR_UNSIGNED__ that prevents it
being redefined if it’s defined already, so checking $GCC at
configure time is redundant and can safely be removed.
* lib/autoconf/c.m4 (AC_C_CHAR_UNSIGNED): Do not make result depend on
value of $GCC. Adjust commentary.
If an error message starts with an entire fully uppercased word,
that’s probably a proper noun and it should stay that way. For
instance, autoreconf has an error message that starts with
"AM_GNU_GETTEXT is used, but ..."; AM_GNU_GETTEXT is the name of an
Automake macro, it needs to stay uppercased.
This subsumes the existing exception for the words FATAL and WARNING.
While I was in there I generalized the exception for PRIuMAX to cover
all of the inttypes.h PRI* and SCN* macros.
This patch has been submitted to Gnulib; until it is merged there,
anyone running ‘make fetch’ should take care not to drop this change.
* maint.mk (sc_error_message_uppercase): Allow error messages that
begin with any fully uppercased word, or with any of the inttypes.h
PRI[dioux]\w+ or SCN[dioux]\w+ macros.
Several of the files updated by make fetch have made this change, so
we follow suit for consistency:
* README-hacking
* build-aux/fetch.pl
* doc/autoconf.texi: Replace all git.savannah.gnu.org/gitweb URLs with
equivalent /cgit URLs.
Fix some subtle quotation bugs in _AC_INIT_PACKAGE that made it
impossible to put ‘,’ or an unbalanced close parenthesis in some of
the arguments to AC_INIT. Document that arguments to AC_INIT
containing parentheses, square brackets, ‘,’ or ‘#’ may need to be
double-quoted. Provide more detailed examples and exposition re
computing the arguments to AC_INIT when autoconf is run (e.g. with
git-version-gen). Add a whole bunch more tests for unusual arguments
to AC_INIT, and a test that the backward-compatibility behavior of
AC_INIT with only one argument is still correct.
This may still break some of the existing configure scripts described
in the threads at
https://lists.gnu.org/r/autoconf/2020-10/msg00013.html and
https://lists.gnu.org/r/bug-autoconf/2020-10/msg00012.html
but, I hope, only in ways covered by the existing warning in NEWS
about pickier M4 quotation.
* lib/autoconf/general.m4 (_AC_INIT_PACKAGE): Redo argument
normalization and default value selection in a simpler, less
error-prone fashion.
(_AC_INIT_PACKAGE_N): New helper subroutine.
(AC_INIT): Always call _AC_INIT_PACKAGE, but supply no arguments if
we were called with only one argument.
* tests/base.at (AC_INIT (obsolete invocation)): New test.
(AC_INIT with unusual version strings): Expand test.
* doc/autoconf.texi (AC_INIT): Revise.
AS_ECHO expands to ‘printf "%s\n" $1’. If a configure script defines
an M4 macro named ‘s’ or ‘n’ it will be expanded in the first argument
to printf, which is almost certainly not what was intended.
The configure script for ruby 2.7.2 uses ‘AS_VAR_PUSHDEF([s], ...)’
and breaks with 2.69d because of this.
Add some extra quoting so that the ‘%s\n’ is treated as literal;
similarly for AS_ECHO_N and the legacy shell variables $as_echo
and $as_echo_n.
For now, anyway, don’t quote the word ‘printf’; if someone does
define that as a M4 macro they might well mean to affect AS_ECHO.
(Whether this is something we *want* to allow, we can worry about
when it comes up.)
Fixes bug #110377.
* lib/m4sugar/m4sh.m4 (_AS_ECHO_N_PREPARE, AS_ECHO, AS_ECHO_N):
Add another layer of quoting around the first argument to printf.
* tests/m4sh.at (Redefining AS_ECHO internals): New test.
The most recently released version of libtool.m4 is five years old as
of this commit, and no new release is likely to appear anytime soon.
It uses AC_LANG_SAVE and AC_LANG_RESTORE, in a way that doesn’t
obviously translate to AC_LANG_PUSH and AC_LANG_POP. This will need
to be fixed by libtool upstream. Until that actually happens, disable
the -Wobsolete warnings for AC_LANG_SAVE and AC_LANG_RESTORE. (They
are still documented as obsolete in the manual, as they have been for
many years.)
Fixes bug #110375.
* lib/autoconf/lang.m4 (AC_LANG_SAVE, AC_LANG_RESTORE): Define with
AC_DEFUN, not AU_DEFUN; remove manual -Wobsolete warnings.
AS_IF can emit a syntactically invalid shell if-then-else,
if CONDITION
then :
# ...
else
fi
when its IF-FALSE argument consists of macros that don’t produce any
shell code. This was a documented limitation in AS_IF, but it’s a bad
limitation to have, because macros that *used* to expand to shell
commands might start expanding to nothing in future releases. For
instance, this broke the libzmq configure script, which did
AC_PROG_CC
AX_CHECK_COMPILE_FLAG([-std=gnu11],
[CFLAGS+=" -std=gnu11"],
[AC_PROG_CC_C99])
Perfectly valid in 2.69, but in 2.70 AC_PROG_CC_C99 doesn’t produce
any shell code and the script crashes.
We had that limitation for good reason: we can’t just put ‘:’ at the
beginning of the else-clause, like we do for the then-clause, because
that would clobber $? and the IF-FALSE commands might want to inspect
it. (This doesn’t matter for the then-clause, because $? is always
zero at the beginning of a then-clause anyway.) The simplest and
least inefficient shell construct I can find that works in this
context is a shell function that does ‘return $?’. Due to awkward
M4sh initialization ordering constraints (AS_IF gets used before we
can safely use shell functions) an indirection through a shell
variable is necessary. The structure of a m4sh script is now
#! /bin/sh
## M4sh Initialization
as_nop=:
...
## M4sh Shell Functions
as_fn_nop () { return $?; }
as_nop=as_fn_nop
...
and AS_IF emits
if CONDITION
then :
# ...
else $as_nop
# ...
fi
The uses of AS_IF that appear before the beginning of the M4sh Shell
Functions section are all under our control and they don’t need to
look at $?.
If anyone has a better idea for how to make this work I will be glad
to hear it.
Fixes bug #110369.
* lib/m4sugar/m4sh.m4
(_AS_IF_ELSE): When $1 is nonempty, invoke _AS_EMPTY_ELSE_PREPARE.
Emit $as_nop at beginning of else clause.
(_AS_BOURNE_COMPATIBLE): Initialize as_nop to ‘:’.
(_AS_EMPTY_ELSE_PREPARE): New macro which emits a definition of
as_fn_nop and resets as_nop to as_fn_nop.
(AS_PREPARE, _AS_PREPARE): Invoke _AS_EMPTY_ELSE_PREPARE.
(_AS_UNSET_PREPARE): Tweak white space.
* tests/m4sh.at (AS_IF and AS_CASE): Test AS_IF’s IF-FALSE argument
being empty after macro expansion.
* doc/autoconf.texi (AS_IF): Remove warning about use with
‘run-if-false’ argument empty after macro expansion.
Instead of treating CONFIG_SITE as a single path, treat it as a
space-separated list of paths and load them in order.
Also remove the special-casing of entries starting with a dash, this is
redundant as they'll be caught by the wildcard case.
Finally add a test case to verify that multiple files are loaded
correctly.
* lib/autoconf/general.m4 (AC_SITE_LOAD): Treat CONFIG_SITE as a
space-separated list of scripts to be sourced. Simplify handling
of default config.site locations using this capability.
* tests/base.at (AC_CACHE_CHECK): Test loading of multiple site files.
* doc/autoconf.texi (Site Defaults): Update documentation of CONFIG_SITE.
In 2003, Joey Hess reported the following bug against Debian's
autoconf package (see http://bugs.debian.org/221483):
I noticed that if I ctrl-c autoconf, it can leave a partially
written, executable configure script. I was lucky enough to
get a configure script that exited with a shell parse error,
but if I had been unlucky, it might have exited 0 without
doing all the tests I expected it to do. That would have
sucked to ship to users.
There are many ways to update a file in a way that is not
prone to these problems, and I suggest that autoconf adopt
one of them.
Ben Pfaff wrote a patch to make autom4te replace the output file
atomically; Debian has carried it since 2006. He submitted it
to autoconf upstream in 2008 but it never went anywhere.
I (Zack) have dusted off the patch and made some minor improvements:
using File::Temp (with DIR set to the directory of the output file)
instead of a predictable temporary file name, and using
Autom4te::FileUtils::update_file instead of File::Copy::move.
I do not attempt to test the fix (the test would be inherently racey)
nor do I have autom4te delete the temp file if it crashes while the
file is being written (there is no way to do this with 100%
reliability and it strikes me as likely to cause more problems than it
solves).
Fixes our bug #110305.
* bin/autom4te.in (handle_output): When $output is to a regular or
nonexistent file, write to a temporary file in the same directory
and then rename it over $output after completion.
As of the 2020-11-07 update, config.sub and config.guess
unconditionally use $(...) command substitution; see
<https://lists.gnu.org/archive/html/config-patches/2020-11/msg00011.html>.
Therefore, add this to the set of required shell features, searched
for by _AS_DETECT_BETTER_SHELL. On a system where /bin/sh doesn’t
support $(...), $CONFIG_SHELL will be set to one that does (and the
primary configure script will be re-executed using that shell).
AC_CANONICAL_* use $CONFIG_SHELL to execute config.guess/sub, so they
will keep working. This also means that configure scripts and
third-party macros that use $(...) will quietly start working
correctly on such ancient systems.
The test code is simple, but sufficient to weed out Solaris 10’s
/bin/sh, which doesn’t support $(...) but *does* support shell
functions.
I’m not going to touch any of the existing uses of `...` command
substitution in Autoconf proper for now, but it might make sense to
bulk upgrade them early in the 2.71 release cycle; if nothing else,
it would remove a major obstacle to running shellcheck over our
scripts.
* lib/m4sugar/m4sh.m4 (_AS_MODERN_CMDSUBST_WORKS): New macro.
(AS_INIT, AS_SHELL_SANITIZE): Call _AS_DETECT_REQUIRED for
_AS_MODERN_CMDSUBST_WORKS.
* NEWS: Mention the requirement for $(...).
As pointed out in sr #110368, since install-sh is now being installed
as part of autoconf, we should make sure to ship the latest version
rather than the version shipped by the automake that was used to
bootstrap the autoconf release tarball.
The build-aux/fetch.pl script is already supposed to fetch the latest
version, but install-sh is listed in .gitignore so any updates are
discarded when starting from a clean tree. Correct this.
At the same time, since mdate-sh is *not* installed by autoconf nor is
it directly referenced in any code maintained in the autoconf
repository, remove it from the list of files to fetch and keep it in
.gitignore.
This change exposed a bug in fetch.pl where it would crash when
there was no old copy of a file being updated.
* .gitignore: Remove /build-aux/install-sh.
* build-aux/fetch.pl (%to_fetch): Remove build-aux/mdate-sh.
(slurp): Don’t die on ENOENT, return undef.
(replace_if_change): Handle $oldcontents being undef.
* build-aux/install-sh: Is now checked in.
As reported in https://github.com/gentoo/gentoo/pull/163, gettext 0.19.6
supports using AM_GNU_GETTEXT_REQUIRE_VERSION (specifying a *minimum*
required version of gettext) instead of AM_GNU_GETTEXT_VERSION
(which specifies a *fixed* required version of gettext).
Update autoreconf to support both.
* bin/autoreconf.in (autoreconf_current_directory): Check for
AM_GNU_GETTEXT_REQUIRE_VERSION as well as AM_GNU_GETTEXT_VERSION.
Update diagnostics about using AM_GNU_GETTEXT_VERSION but not
AM_GNU_GETTEXT, or vice versa, to match.
* doc/autoconf.texi (autoreconf Invocation): Update to match.
Found by exhaustive testing for differences between probe results
under AC_LANG(C) and AC_LANG(C++).
* lib/autoconf/c.m4 (AC_C_FLEXIBLE_ARRAY_MEMBER): Cast result of
malloc for C++ compatibility.
* lib/autoconf/programs.m4 (_AC_PROG_LEX_YYTEXT_DECL): Declare yywrap
as extern "C" when compiling as C++.
While testing the previous patch I noticed that the ‘forbidden tokens,
basic’ test can fail if it runs too fast, because the autom4te cache
files aren’t considered newer than configure.ac.
* tests/tools.at (forbidden tokens, basic): Add delays to ensure
autom4te.cache files are newer than configure.ac.
autoreconf expects to find $(pkgdatadir)/build-aux/config.sub etc
under those names, not names modified by --program-transform-name.
Placing them in $(pkgdatadir) is sufficient to keep parallel
installations of autoconf separate: anyone doing that would need
to adjust @PACKAGE@ anyway.
* lib/local.mk: Use a _DATA rule, not a _SCRIPTS rule, to install
config.guess, config.sub, and install-sh.
(install-data-hook-make-aux-scripts-executable): New hook rule.
I misremembered how AC_LANG_PROGRAM works. We don’t need to invoke
AC_INCLUDES_DEFAULT here but we *do* need to explicitly include
string.h.
Unfortunately we have no good way of testing for this regression with
the testsuite as it is today.
* lib/autoconf/functions.m4 (AC_FUNC_STRERROR_R): Include string.h in
test program.
Some widely used Automake recipes involve putting AC_REQUIRE_AUX_FILE
at top level of a configure script, and it uses AC_REQUIRE now, so it
needs to be defined with AC_DEFUN.
* lib/autoconf/general.m4 (AC_REQUIRE_AUX_FILE): Define with AC_DEFUN.
* tests/torture.at (Missing auxiliary files (foreign)): New test.
Some versions of lex need you to forward-declare yywrap in a %{ %}
block before the rules section, if you’re going to define it yourself.
May help with bug #110312.
* lib/autoconf/programs.m4 (_AC_PROG_LEX_YYTEXT_DECL): In the test
input to lex, forward-declare yywrap before the rules.
In very old perls (I noticed this with 5.8.4), File::Temp objects are
not automatically stringified in all contexts where we expect them to
be, causing autoreconf to crash.
* bin/autoreconf.in (install_aux_file): Explicitly extract the
temporary file’s name from $temp, and use that in all the places we
were using $temp.
Commit 29ede6b96f caused AC_PROG_LEX to
stop looking for a library that provides yywrap. This broke several
packages in a Debian archive rebuild.
Revert all the way to the 2.69 behavior, which was to set LEXLIB to
-ll or -lfl if that library defines yywrap, but allow AC_PROG_LEX to
succeed if neither -ll nor -lfl exists on the system, even if a lex
program that doesn't define yywrap would need it.
(This behavior was a bug, but people have come to depend on it.
See https://savannah.gnu.org/support/index.php?110269 and the
thread starting from
https://lists.gnu.org/r/autoconf-patches/2020-07/msg00013.html
for gory details.)
To provide a path away from bug-compatibility, AC_PROG_LEX now takes
one argument, documented as a whitespace-separated list of options.
Two options are defined: ‘yywrap’ means to look for yywrap and behave
as if lex is unavailable if it isn’t found; ‘noyywrap’ means to not
look for yywrap at all. These are mutually exclusive.
Fixes bug #110346.
* lib/autoconf/programs.m4 (AC_PROG_LEX): Add an argument which
can be either ‘yywrap’, meaning to look for yywrap in -ll, or
‘noyywrap’, meaning to not look for yywrap at all. In the
absence of either option, issue an obsoletion warning and
revert to behavior bug-compatible with 2.69.
* tests/semantics.at: Add more tests of AC_PROG_LEX.
* tests/mktests.sh: Exclude AC_PROG_LEX from autogenerated tests.
* doc/autoconf.texi: Update documentation of AC_PROG_LEX.
* NEWS: Update notes on AC_PROG_LEX.
Some of the compiler options that AC_OPENMP tests, mean “enable
OpenMP” to one compiler, but “write output to a file named ‘mp’ or
‘penmp’” to other compilers. The author of AC_OPENMP believed that
this could only happen if compilation was *successful*, but didn’t
realize that one of the options means “write *preprocessed* output to
a file named ‘penmp’” to SunPRO C, and that this *would* succeed on
the test program. (AC_LINK_IFELSE fails anyway, because the
compilation didn’t create conftest$exeext.)
The option that actually means “enable OpenMP” to SunPRO C is earlier
in the list than the option that means “write preprocessed output to a
file named ‘penmp’”, so we might never have noticed this, but for a
second bug: if you have a bad combination of Solaris operating system
patches installed, it’s possible for this compiler to
successfully *compile* a program that uses OpenMP, but then fail
to *link* it because the OpenMP runtime library is out of sync with
the core C library. AC_OPENMP doesn’t distinguish this case from
“that option doesn’t mean ‘enable OpenMP’” so it goes on to other
entries in the list and hits the “write preprocessed output” one.
Implement four layers of defensive measures against this mess:
- Use an #error directive instead of a compile-time syntax error
to halt compilation when _OPENMP is not defined.
- For each option that might mean “enable OpenMP”, first do an
AC_COMPILE_IFELSE to find out whether it really means that, and
then an AC_LINK_IFELSE to find out whether it works. If the
compilation succeeds but the link fails, bail out of the loop and
declare OpenMP to be unsupported.
- If a file named ‘mp’ or ‘openmp’ exists in configure’s working
directory when AC_OPENMP begins, error out. This means it is safe
to delete any file named ‘mp’ or ‘openmp’ that exists at the *end*
of AC_OPENMP.
- If a file named ‘mp’ or ‘openmp’ exists in the top level of the
source tree with a configure.ac that uses AC_OPENMP, have autoconf
error out, too.
Fixes bug #110353. Problem reported by Dagobert Michelsen.
* lib/autoconf/c.m4 (_AC_LANG_OPENMP(C)): Change ‘choke me’ to
‘#error "OpenMP not supported"’.
(AC_OPENMP): AC_REQUIRE _AC_OPENMP_SAFE_WD. For each option, do
both a compile test and a link test; if the compile test succeeds
but the link fails, don’t go on to other candidate options.
Delete files named ‘mp’ and ‘penmp’ after the loop.
(_AC_OPENMP_SAFE_WD): New macro, subroutine of AC_OPENMP. If files
named ‘mp’ or ‘penmp’ exist, error out both at autoconf time and at
configure time.
* tests/torture.at (Files clobbered by AC_OPENMP): New test.
* doc/autoconf.texi: Document requirement not to have files
named ‘mp’ or ‘penmp’ next to a configure.ac that uses AC_OPENMP.
Several tests in the testsuite run a system-provided aclocal, which
will look into its $prefix/share/aclocal for third-party macros.
If those macros are buggy, aclocal may bomb out even though the test
doesn’t use them, causing the test to fail spuriously.
In all tests that need to run aclocal, create an empty directory and
give aclocal the --system-acdir option pointing at that directory.
This masks out all these potentially buggy macros. (It does *not*
mask out AM_INIT_AUTOMAKE, which aclocal will find in a different
directory.)
In all tests that run autoreconf but *don’t* need to run aclocal,
create an empty aclocal.m4 and set ACLOCAL=true in the environment.
Fixes bug #110352. Problem reported by Dagobert Michelsen.
* tests/fortran.at
* tests/tools.at
* tests/torture.at:
Set ACLOCAL=true in the environment in all tests that run
autoreconf but don’t need to run aclocal.
Set ACLOCAL="aclocal --system-acdir <empty directory>" in all
tests that do need to run aclocal.
