Bruno Haible [Tue, 11 Dec 2018 19:36:35 +0000 (20:36 +0100)]
array-set, linkedhash-set, hash-set: Fix tests.
* tests/test-array_set.c (cmp_objects_in_array): New function.
(check_equals): Use it.
* tests/test-hash_set.c: Likewise.
* tests/test-linkedhash_set.c: Likewise.
Bruno Haible [Mon, 3 Dec 2018 23:53:12 +0000 (00:53 +0100)]
linkedhash-set: New module.
* lib/gl_linkedhash_set.h: New file.
* lib/gl_linkedhash_set.c: New file.
* lib/gl_anyhash_set1.h: New file, based on lib/gl_anyhash_list1.h.
* lib/gl_anyhash_set2.h: New file, based on lib/gl_anyhash_list2.h.
* lib/gl_anyhash_primes.h: New file, extracted from
lib/gl_anyhash_list2.h.
* lib/gl_anyhash_list2.h: Include it.
(primes, next_prime): Remove definitions.
* modules/linkedhash-set: New file.
* modules/avltreehash-list (Files): Add lib/gl_anyhash_primes.h.
(Makefile.am): Add gl_anyhash_primes.h to lib_SOURCES.
* modules/linkedhash-list (Files): Add lib/gl_anyhash_primes.h.
(Makefile.am): Add gl_anyhash_primes.h to lib_SOURCES.
* modules/rbtreehash-list (Files): Add lib/gl_anyhash_primes.h.
(Makefile.am): Add gl_anyhash_primes.h to lib_SOURCES.
Akim Demaille [Fri, 7 Dec 2018 05:26:08 +0000 (06:26 +0100)]
bison: don't force the Yacc mode
Passing -y forces Bison into POSIX YACC mode. This includes reporting
errors when Bison features are used in the grammar file. Some of
these features (such as %expect) were flagged non-yacc recently. Most
of the time, -y is actually used to please Automake's ylwrap which
expects the output to be y.tab.c.
* m4/bison.m4 (gl_BISON): Use `-o y.tab.c` rather than `-y`.
Eric Blake [Thu, 29 Nov 2018 19:33:28 +0000 (13:33 -0600)]
docs: mention printf %m considerations
gcc -Wformat does NOT flag printf("%m") on a BSD system, even though
it prints only "m" instead of the desired GNU/Linux behavior of the
current errno value (you have to use -Wpedantic to flag it, but that's
a rather heavy hammer). Still, it's easy enough (even if verbose)
to write out strerror() usage directly, to the point that gnulib will
probably never provide a module to guarantee %m everywhere.
See also https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88270
Paul Eggert [Thu, 29 Nov 2018 00:10:03 +0000 (16:10 -0800)]
strerror_r-posix: memmove, not memcpy
* lib/strerror_r.c (safe_copy): Use memmove, not memcpy,
since the source and destination might overlap in the call
‘safe_copy (buf, buflen, strerror_r (errnum, buf, buflen))’.
Simplify.
Paul Eggert [Fri, 2 Nov 2018 21:09:35 +0000 (14:09 -0700)]
gnulib-common.m4: port _Noreturn to C++
Problem reported by Akim Demaille in:
https://lists.gnu.org/r/bug-bison/2018-10/msg00067.html
* m4/gnulib-common.m4 (gl_COMMON_BODY): If C++, use [[noreturn]].
Merge adjustments from _Noreturn.h and from glibc into the non-C++
version.
* lib/_Noreturn.h: Match gnulib-common.
Norihiro Tanaka [Mon, 22 Oct 2018 14:51:20 +0000 (23:51 +0900)]
dfa: a state has a set of current positions.
Up to now, a state had a set of follow-on positions. It is replaced a
set of current positions. This change will save memory space.
dfa.c (leaf_set): Remove it.
(struct dfa): Add new member constraints and separates.
(append): New function.
(state_index): Bring constraint from pre-calculated.
(state_separate_contexts): Bring separate contexts from pre-calculated.
Change argument, update callers.
(merge_nfa_state): Pre-calculate constraints for END. and remove END.
No longer END is not used after here.
(dfaoptimize): Initialize added member constraints.
(dfaanalyze): Pre-calculate seprate contexts.
(build_state): Change for this update.
(dfassbuild): Initialize new members .
(dfafree): Free memory for new members.
Norihiro Tanaka [Mon, 22 Oct 2018 14:31:26 +0000 (23:31 +0900)]
dfa: position set sorts increasing order
Change the order of position set from decreasing to increaing, then even
after dfa is optimized, it is guaranteed that the number of a position
is smaller than the subsequent one's number.
dfa.c (insert, merged_constrained, delete): Reverse the direction of an
inequality sign.
(dfaanalyze): Position set sorts increasing order.
Paul Eggert [Thu, 25 Oct 2018 15:33:59 +0000 (08:33 -0700)]
havelib: fix nested ‘configure’ chatter
* m4/lib-prefix.m4 (AC_LIB_PREPARE_MULTILIB): Don’t nest
AC_CACHE_CHECK calls, which resulted in confusing output like
“checking for the common suffixes of directories in the library
search path... checking for 64-bit host... no lib,lib”.
Paul Eggert [Thu, 25 Oct 2018 15:25:41 +0000 (08:25 -0700)]
backupfile: tweak for better code
* lib/backupfile.c: Sort include directives, and remove
unnecessary <limits.h> include.
(FALLTHROUGH): New macro, copied from other modules.
(backupfile_internal): Use it to avoid code duplication.
This lets GCC 8.2.1 generate better code by inlining the
call to check_extension.
Paul Eggert [Wed, 24 Oct 2018 02:10:21 +0000 (19:10 -0700)]
backupfile: new dir_fd args
New module opendirat with code taken from fts.
Use this module to let backupfile use a directory file descriptor.
* NEWS: Document the incompatible change.
* lib/backup-find.c (find_backup_file_name):
* lib/backup-rename.c (backup_file_rename):
New arg DIR_FD.
* lib/backupfile.c: Include stdint.h, for SIZE_MAX.
(SIZE_MAX): Remove.
Include opendirat.h rather than dirent--.h.
(check_extension): New args DIR_FD and BASE_MAX. All callers changed.
(numbered_backup): New args DIR_FD and PNEW_FD. All callers changed.
(backupfile_internal): New arg DIR_FD. All callers changed.
* lib/fts.c: Include opendirat.h.
(opendirat): Move to opendirat.c.
* lib/opendirat.c, lib/opendirat.h, modules/opendirat: New files.
* modules/backupfile (Depends-on): Remove dirfd, opendir.
Add opendirat.
* modules/fts (Depends-on): Remove fdopendir, openat-safer.
Add opendirat.
Bruno Haible [Mon, 22 Oct 2018 22:06:52 +0000 (00:06 +0200)]
Assume Autoconf >= 2.63.
* DEPENDENCIES: Mention the requirement.
* gnulib-tool (DEFAULT_AUTOCONF_MINVERSION): Bump to 2.63.
(func_get_filelist): Don't list m4/onceonly.m4 any more.
* pygnulib/GLModuleSystem.py (getFiles): Likewise.
* m4/onceonly.m4: Remove file.
Bruno Haible [Mon, 22 Oct 2018 00:38:39 +0000 (02:38 +0200)]
localename: Fine-tune support for per-thread locales on Solaris 11.4.
* lib/localename-table.h: New file, extracted from lib/localename.c.
* lib/localename-table.c: Likewise.
* lib/localename.c: Include localename-table.h.
(get_locale_t_name, newlocale, duplocale, freelocale): Invoke
locale_hash_function instead of pointer_hash.
* modules/localename (Files): Add lib/localename-table.h,
lib/localename-table.c.
(lib_SOURCES): Add localename-table.c.
* m4/intlsolaris.m4 (gt_INTL_SOLARIS): Require AC_CANONICAL_HOST. Test
for Solaris 11.4 locale system only on Solaris. Test for it
independently whether getlocalename_l exists.
* m4/intl.m4 (gt_INTL_SUBDIR_CORE): Don't test for 'uselocale' and
'getlocalename_l'. Instead, invoke gt_INTL_SOLARIS. Set
HAVE_NAMELESS_LOCALES.
* modules/gettext (Files): Add m4/intlsolaris.m4.
Bruno Haible [Mon, 22 Oct 2018 00:34:07 +0000 (02:34 +0200)]
Small update from gettext.
* m4/intl.m4: Update from gettext:
- 2018-01-02: Fix 'ar' invocation when cross-compiling and in 64-bit
mode on AIX.
- 2018-01-02: Don't use -lc explicitly when linking with libtool.
- 2017-05-19: (AM_INTL_SUBDIR): Require AC_C_FLEXIBLE_ARRAY_MEMBER.
Akim Demaille [Fri, 12 Oct 2018 04:46:09 +0000 (06:46 +0200)]
timevar: use gethrxtime to get wall clock time
clock_gettime is not portable. gethrxtime takes the best available
option to get the wall clock time, including clock_gettime (monotonic
clock), and gettime (non monotonic).
Also, using xtime_t instead of float preserves the precision.
Suggested by Bruno Haible.
* lib/xtime.h (xtime_make): Handle overflows of nanoseconds.
* modules/timevar (Depends-on): We need gethrxtime.
We no longer use times().
(Link): Update.
* lib/timevar.h (timevar_time_def): Use xtime_t.
* lib/timevar.c (set_to_current_time): Use gethrxtime.
(timevar_print): Instead of checking whether the timings themselves
are large enough for the timevar to be printed, check the percentages.
Bruno Haible [Sun, 14 Oct 2018 07:21:24 +0000 (09:21 +0200)]
wcsnrtombs: Work around Solaris 11.4 bug.
* m4/wcsnrtombs.m4 (gl_WCSNRTOMBS_WORKS_IN_TRADITIONAL_LOCALE): New
macro.
(gl_FUNC_WCSNRTOMBS): Invoke it.
* doc/posix-functions/wcsnrtombs.texi: Mention the Solaris bug.
Bruno Haible [Sun, 14 Oct 2018 07:19:12 +0000 (09:19 +0200)]
mbsnrtowcs: Work around Solaris 11.4 bug.
* m4/mbsnrtowcs.m4 (gl_MBSNRTOWCS_WORKS_IN_TRADITIONAL_LOCALE): New
macro.
(gl_FUNC_MBSNRTOWCS): Invoke it.
* doc/posix-functions/mbsnrtowcs.texi: Mention the Solaris bug.
* m4/floor.m4 (gl_FUNC_FLOOR_LIBS): Use 'floor' also through a function
pointer.
* m4/ceil.m4 (gl_FUNC_CEIL_LIBS): Use 'ceil' also through a function
pointer.
* m4/trunc.m4 (gl_FUNC_TRUNC): Use 'trunc' also through a function
pointer.
* m4/truncf.m4 (gl_FUNC_TRUNCF): Use 'truncf' also through a function
pointer.
* m4/truncl.m4 (gl_FUNC_TRUNCL): Use 'truncl' also through a function
pointer.
Akim Demaille [Thu, 11 Oct 2018 15:54:35 +0000 (17:54 +0200)]
timevar: expect that getrusage is available
Don't keep both times and getrusage as backend: both are guaranteed by
gnulib, a single one suffices. Using getrusage is open to possibly
tracking other types of resources in the future.
* modules/timevar (Depends-on): Add getrusage.
(configure.ac): Remove gl_TIMEVAR.
(Files): Remove m4/timevar.m4.
* m4/timevar.m4: Remove, rely on gnulib for getrusage.
* lib/timevar.h (timevar_enabled): Clarify documentation.
* lib/timevar.c: Remove all the code about times.
Remove all the CPP guards about getrusage: expect it to be present
(courtesy of gnulib).
Bruno Haible [Fri, 12 Oct 2018 09:12:53 +0000 (11:12 +0200)]
mountlist: Improve support for Solaris in 64-bit mode.
Reported by David Wood <David.Wood@deshaw.com> in
<https://debbugs.gnu.org/cgi/bugreport.cgi?bug=6816>.
* m4/ls-mntd-fs.m4 (gl_LIST_MOUNTED_FILE_SYSTEMS): On Solaris 8 or
newer, define MOUNTED_GETEXTMNTENT instead of MOUNTED_GETMNTENT2.
* lib/mountlist.c: Add code for MOUNTED_GETEXTMNTENT case.
Paul Eggert [Mon, 8 Oct 2018 23:53:59 +0000 (16:53 -0700)]
fts: cleanup after FTS_NOATIME removal
* lib/fts_.h (FTS_VERBATIM, FTS_OPTIONMASK, FTS_NAMEONLY)
(FTS_STOP): Shrink to minimal values. We don’t need to
worry about binary compatibility in Gnulib, and the old way
of doing things had a hole in the user options that caused
FTS_OPTIONMASK to not work as desired.
This reverts commit da4d6974013c822af1498941e32db774b2031765.
We cannot guarantee that O_NOATIME works: e.g. openat fails
with EPERM if the effective user ID of the caller does not match
the owner of the file and the caller is not privileged.
Downstream findutils has never picked up FTS_NOATIME. Discussed at
<https://lists.gnu.org/r/bug-gnulib/2018-09/msg00122.html>.
* lib/fts_.h (FTS_NOATIME): Remove bit flag.
(FTS_OPTIONMASK): Adjust.
* lib/fts.c (diropen, fts_open, fts_build): Likewise.
(fd_ring_check): Likewise.