Paul Eggert [Sun, 20 Nov 2011 22:08:33 +0000 (14:08 -0800)]
port to GNU hosts, where getuid and friends can fail
* src/groups.c (main):
* src/install.c (need_copy):
* src/su.c (log_su):
* src/test.c (unary_operator):
* src/whoami.c (main):
Don't assume that getuid and friends always succeed.
This fixes the same problem that we recently fixed with 'id'.
Jim Meyering [Sun, 20 Nov 2011 14:34:59 +0000 (15:34 +0100)]
maint: avoid gitlog-to-changelog diagnostic about unused entries
* build-aux/git-log-fix: Comment out two unused entries.
Each of those two entries does indicate an error in a commit log,
but precedes the cut-off date, so has an actual VC'd ChangeLog entry.
I.e., gitlog-to-changelog generates ChangeLog entries since 2008-02,
and these two predate that.
* ChangeLog-2008: Make the indicated correction.
Erik Auerswald [Sat, 19 Nov 2011 21:07:29 +0000 (22:07 +0100)]
ln: fix position of --backup values description
* src/ln.c (usage): A paragraph describing interactions of -s
with -L and -P somehow snuck in between the description of the
--backup option and the values used to control it. Fix this by
moving the value description up.
Jim Meyering [Fri, 18 Nov 2011 14:33:56 +0000 (15:33 +0100)]
scripts: rewrite git commit-msg hook in Perl
* scripts/git-hooks/commit-msg: Rewrite in perl.
This is still a work in progress in that it hard-codes coreutils-
specific program names and policies that should be easy to selectively
enable or disable without modifying the script.
Eric Blake [Wed, 16 Nov 2011 17:09:12 +0000 (10:09 -0700)]
sort: clarify wording on -k syntax
* src/sort.c (usage): Use KEYDEF instead of POS, and call out the
specific OPTS that can occur in KEYDEF.
Based on a report by Lars Noodén, http://bugs.gnu.org/10019
Paul Eggert [Tue, 15 Nov 2011 21:23:24 +0000 (13:23 -0800)]
id: fix bug when euid != ruid
* src/id.c (main): Report an error if no args are given and getuid
fails, because print_full_info needs ruid. Redo code so that
getuid and friends are invoked only when needed; this makes the
code easier to follow, and is how I found the above bug.
Ruediger Meier [Tue, 15 Nov 2011 18:55:23 +0000 (11:55 -0700)]
ls: another reword of generic size note
* src/system.h (emit_size_note): Use "unit" rather than "suffix",
and move multiplication to example instead of in suffix list.
See additional discussion in Bug#9939.
Paul Eggert [Tue, 15 Nov 2011 17:30:49 +0000 (09:30 -0800)]
id: handle (uid_t) -1 more portably
* src/id.c (GETID_MAY_FAIL): Remove.
(main): Check for nonzero errno, rather than having a compile-time
GETID_MAY_FAIL guess. Suggested by Roland McGrath in
<http://debbugs.gnu.org/cgi/bugreport.cgi?bug=10021#47>.
Also, the old code was incorrect if uid_t was narrower than int.
(print_full_info): Remove unnecessary cast to -1.
Paul Eggert [Tue, 15 Nov 2011 16:20:59 +0000 (08:20 -0800)]
doc: update for ISO/IEC 80000-13
* doc/coreutils.texi (Block size): IEC 60027-2 has been superseded
by ISO/IEC 80000-13, so prefer the newer standard but also mention
the old. The new standard specifies Zi and Yi, so they are no
longer GNU extensions. Fix stale URL to BIPM.
2011-11-14 Paul Eggert <eggert@cs.ucla.edu>
id: handle (uid_t) -1 more portably
* src/id.c (GETID_MAY_FAIL): Remove.
(main): Check for negative return values, not for -1.
The old code was incorrect if uid_t was narrower than int,
regardless of whether we were on a GNU or a POSIX platform.
The new code is simpler and doesn't need GETID_MAY_FAIL.
(print_full_info): Remove unnecessary cast to -1.
Ludovic Courtès [Sat, 12 Nov 2011 00:25:45 +0000 (01:25 +0100)]
id: fail when getuid, getgid, etc. fail, e.g., on GNU/Hurd
POSIX-conforming getuid, geteuid, etc. functions cannot fail,
but on GNU/Hurd systems and some others, they may.
* src/id.c (main) [__GNU__]: Detect and diagnose any such failure.
* tests/id/gnu-zero-uids: New file.
* tests/Makefile.am (TESTS): Add it to the list.
* tests/init.cfg (require_gnu_): New function.
Jim Meyering [Sat, 12 Nov 2011 21:23:09 +0000 (22:23 +0100)]
maint: add rule to ensure that our commit hook copies are up to date
* Makefile.am (check-git-hook-script-sync): New rule -- not used
anywhere, because it depends on having very recent git.
* scripts/git-hooks/pre-applypatch: New file.
Jim Meyering [Thu, 10 Nov 2011 08:27:17 +0000 (09:27 +0100)]
maint: adjust the URL that will appear in the generated announcement
* cfg.mk (url_dir_list): Use this http://ftp.gnu.org/gnu/$(PACKAGE)
for the first link listed in the generated announcement.
announce-gen now provides the faster mirror link automatically.
Jim Meyering [Tue, 8 Nov 2011 18:03:39 +0000 (19:03 +0100)]
ls: plug a per-argument leak
Using ls -l on an SELinux-enabled system would leak one SELinux
context string per non-empty-directory command-line argument.
* src/ls.c (free_ent): New function, factored out of...
(clear_files): ...here. Use it.
(extract_dirs_from_files): Call free_ent (f), rather than simply
free (f->name). The latter failed to free the possibly-malloc'd
linkname and scontext members, and thus could leak one of those
strings per command-line argument.
* THANKS.in: Update.
* NEWS (Bug fixes): Mention it.
Reported by Juraj Marko in http://bugzilla.redhat.com/751974.
Paul Eggert [Sat, 12 Nov 2011 08:20:01 +0000 (00:20 -0800)]
* tests/misc/sort-continue: Port to Fedora 15.
Redirect with the shell command, not in a separate 'exec'.
Without this patch, Fedora 15 x86-64 /bin/sh (i.e., Bash 4.2.10)
complained about running out of file descriptors in the shell.
Paul Eggert [Sat, 12 Nov 2011 07:21:13 +0000 (23:21 -0800)]
ls: -k no longer affects -l's file sizes
This fixes an incompatibility with POSIX 2008 and with BSD.
Problem reported by Abdallah Clark (Bug#9939)
via Alan Curry (Bug#10016).
* NEWS: Document this.
* doc/coreutils.texi (General output formatting): Document the
new -k behavior, and --kibibytes.
* src/ls.c (file_human_output_opts): New static var.
(long_options, usage): Add --kibibytes.
(decode_switches, gobble_file, print_long_format):
Implement the new -k behavior.
* tests/ls/block-size: New file.
* tests/Makefile.am (TESTS): Add it.
Pádraig Brady [Thu, 3 Nov 2011 01:42:43 +0000 (01:42 +0000)]
tests: avoid signal issues in timeout-group
These issues were seen on an OpenSuse 10.3 system
(kernel 2.6.22.5 x86_64, glibc 2.6.1-18, bash updated to 4.2),
and also on a 64 bit SLES system with a 2.6.16 kernel.
Both systems had 2 CPUs.
There were two issues seen. 1. Occasionally the
timeout.cmd shell script would block SIGINT until
the sleep command exited. 2. Much less frequently the
signal handler in the timeout command itself was ignored,
causing SIGALRM to kill the process.
* tests/misc/timeout-group: Detect the above two cases,
and skip rather than fail. Note only issue 2. causes
a failure unless skipped, but we skip for case 1. also,
for diagnostic purposes.
Pádraig Brady [Thu, 13 Oct 2011 13:58:58 +0000 (14:58 +0100)]
tests: fix a race in timeout-group
This could cause a false failure, or even
an infinite loop in rare circumstances.
* tests/misc/timeout-group: Increase the timeouts
passed to the timeout command, so that they're
effectively not used. Instead the command termination
is triggered by the kill commands when everything
is in the correct state.
Jim Meyering [Wed, 2 Nov 2011 09:20:11 +0000 (10:20 +0100)]
build: adapt to gnulib's recent openat/fchmodat separation
* gnulib: Update to latest, pulling in the openat/fchmodat separation.
* bootstrap.conf (gnulib_modules): Add fchmodat, now that gnulib
has moved it into its own module.
Jim Meyering [Mon, 31 Mar 2008 20:39:15 +0000 (22:39 +0200)]
maint: fix git-log typos when generating ChangeLog
* gnulib: Update to latest, for new gitlog-to-changelog.
* Makefile.am (gen-ChangeLog): Use its new --amend=F option.
* build-aux/git-log-fix: New file, with ChangeLog fixes.
Pádraig Brady [Tue, 1 Nov 2011 17:29:53 +0000 (17:29 +0000)]
doc: date: mention that the hardware clock might not be set
* doc/coreutils.texi (Setting the time): Reorganize slightly
and mention that the hardware clock might need to be explicitly
updated by the user as is the case on Fedora 16 currently.
See http://bugzilla.redhat.com/749516
Jim Meyering [Sun, 30 Oct 2011 21:31:17 +0000 (22:31 +0100)]
maint: avoid new "make syntax-check" failure due to stdalign definition
* gl/modules/randread (Depends-on): Add stdalign.
* gl/lib/randread.c: Include <stdalign.h>, so we can ...
[!_STRING_ARCH_unaligned]: remove definition of stdalign.
Jim Meyering [Fri, 28 Oct 2011 16:06:44 +0000 (18:06 +0200)]
tests: don't make tail's pipe-f2 test take the full 10 seconds
* tests/tail-2/pipe-f2: Don't always wait 10 seconds.
Before, this test would always wait 10 seconds.
Now, it stops early when it detects that tail -f has written output.
BTW, the race condition that prompted changing the timeout from 1 second
to 10 was that tail -f could be killed by the timeout before producing
any output.
Jim Meyering [Thu, 27 Oct 2011 20:02:23 +0000 (22:02 +0200)]
date: reinstate the --iso-8601 (-I) option
We deprecated and undocumented the --iso-8601 (-I) option mostly
because date could not parse that particular format. Now that
it can, it's time to restore the documentation.
* src/date.c (usage): Document it.
* doc/coreutils.texi (Options for date): Reinstate documentation.
Reported by Hubert Depesz Lubaczewski in http://bugs.gnu.org/7444.
Jim Meyering [Thu, 27 Oct 2011 17:04:45 +0000 (19:04 +0200)]
build: avoid bootstrap failure when using patch-2.5.8
* gl/modules/tempname.diff: Regenerate to correct bogus offsets
and adjust for 1-line offset. Eric Blake reported that this
patch failed to apply when using patch-2.5.8.
Note that it applies fine using patch-2.6.1.
Jim Meyering [Tue, 25 Oct 2011 10:31:44 +0000 (12:31 +0200)]
build: do not remove definition of ENODATA, after all
* src/system.h (ENODATA): Restore definition.
gnulib defines it only on native Windows systems, so removing our
definition would have provoked build failure on systems that use it,
like FreeBSD. Reported by Bruno Haible in
http://thread.gmane.org/gmane.comp.lib.gnulib.bugs/28739/focus=28795
* gnulib: Update to latest, to get new ENODATA-exempting maint.mk rule.
Jim Meyering [Mon, 24 Oct 2011 08:27:22 +0000 (10:27 +0200)]
rm: update gnulib to get an fts fix for Cygwin+NWFS/NcFsd file systems
* NEWS (Bug fixes): Mention it.
As far as we know, this fix affects only Cygwin with NWFS or NcFsd
file systems. See these:
http://git.sv.gnu.org/cgit/gnulib.git/commit/?id=71f13422f3e634
http://thread.gmane.org/gmane.comp.lib.gnulib.bugs/28739
http://cygwin.com/ml/cygwin/2011-10/msg00365.html
* src/system.h (ENODATA): Remove fall-back definition, now that
gnulib provides one. Caught by the sc_prohibit_always-defined_macros
syntax-check rule.
Also remove now-irrelevant "Don't use bcopy..." comment.
Jim Meyering [Thu, 20 Oct 2011 15:26:01 +0000 (17:26 +0200)]
maint: add git hook scripts
We find it worthwhile to use consistent commit summary prefixes.
To that end, the commit-msg script requires that all commits I make
start with "$P: " (where $P is one of ~100 programs in coreutils)
or one of a few other words, like gnulib tests maint doc build.
It allows more than one word, so e.g., "cat tail head: " would also
be accepted. Pádraig Brady wrote the initial version, with its
72-column and blank-if-present second line checks.
The pre-commit script is the same as the git-supplied sample script,
modulo a bug fix and the "exec 1>&2" redirection.
* scripts/git-hooks/commit-msg: New file.
* scripts/git-hooks/pre-commit: New file.
* scripts/git-hooks/applypatch-msg: New file. Verbatim from .sample.
* cfg.mk: Exempt two of the new scripts from the no-leading-TABs check,
since they're nearly verbatim from git, and we want to stay in sync.
Exempt the commit-msg script from the no-"fail=0" check.
Martial Bornet [Thu, 20 Oct 2011 17:18:09 +0000 (19:18 +0200)]
tail: with -f, use nanosleep (not inotify) on a GPFS file system
Because tail's fremote function did not designate GPFS as
a remote file system type, tail -f would mistakenly attempt
to use inotify, which cannot work with a remote file system.
* src/tail.c (fremote): List GPFS as a remote file system type.
* NEWS (Bug fixes): Mention it.
Reported in http://bugs.debian.org/646022.
Jim Meyering [Tue, 18 Oct 2011 10:04:02 +0000 (12:04 +0200)]
tac: do not let failed allocation cause immediate exit
* src/tac.c (temp_stream): Don't exit immediately upon failed heap
allocation, here. That would inhibit processing of any additional
command-line arguments.
Jim Meyering [Tue, 18 Oct 2011 09:44:39 +0000 (11:44 +0200)]
tac: use only one temporary file, with multiple nonseekable inputs
* src/tac.c (temp_stream): New function, factored out of...
(copy_to_temp): ...here.
(tac_nonseekable): Don't free or fclose, now that we reuse the file.
Suggested by Ambrose Feinstein.
* THANKS.in: Update.
Paul Eggert [Tue, 18 Oct 2011 05:43:58 +0000 (07:43 +0200)]
maint: make tac.c slightly cleaner
* src/tac.c (copy_to_temp): Now that the template string tacXXXXXX
is used in only one place, don't bother using a separate variable.
Also, using three unconditional assignments seems slightly clearer.
Jim Meyering [Sun, 16 Oct 2011 10:07:05 +0000 (12:07 +0200)]
tac: don't misbehave with multiple non-seekable inputs
* src/tac.c (copy_to_temp): Do not reuse the template buffer.
Instead, scribble only on a freshly-xstrdup'd copy each time.
Free that buffer both here, upon failure, and ...
(tac_nonseekable): ...free the buffer in caller, upon success.
* tests/misc/tac-2-nonseekable: New file.
* tests/Makefile.am (TESTS): Add it.
* NEWS (Bug fixes): Mention it.
Reported by Ambrose Feinstein in http://debbugs.gnu.org/9762.
Jim Meyering [Sun, 16 Oct 2011 08:35:56 +0000 (10:35 +0200)]
maint: tac: remove sole use of sprintf in favor of filenamecat
* src/tac.c: Include filenamecat.h.
(copy_to_temp): Use filenamecat rather than xmalloc and sprintf.
Move some declarations "down" to point of initialization.
Pádraig Brady [Mon, 17 Oct 2011 10:20:01 +0000 (11:20 +0100)]
maint: speed up make check (vc_exe_in_TESTS)
* tests/check.mk (vc_exe_in_TESTS): The main change is to
not start a sed process for each file under tests/,
which was taking around 2.5s on a 2.1GHz i3-2310M.
Also adjust the rule to no longer use temporary files.
Jim Meyering [Sun, 9 Oct 2011 08:52:52 +0000 (10:52 +0200)]
rm: do not resort to stat'ing very long names even on deficient systems
This change affects only systems that have neither *at function support
nor the /proc/self/fd support required to emulate those *at functions.
* src/remove.c (write_protected_non_symlink): Call faccessat
unconditionally. Thus we no longer need euidaccess_stat, which was
the sole function used here to operate on a full relative file name.
Remove full_name parameter and update caller.
* lib/euidaccess-stat.h: Remove file.
* lib/euidaccess-stat.c: Likewise.
* m4/euidaccess-stat.m4: Likewise.
* po/POTFILES.in: Remove lib/euidaccess-stat.c.
* m4/prereq.m4 (gl_PREREQ): Don't require gl_EUIDACCESS_STAT.
Prompted by a report from Bruno Haible that the rm/deep-2
test was failing on HP-UX 11.31.
See http://thread.gmane.org/gmane.comp.gnu.coreutils.general/1748
Jim Meyering [Sun, 9 Oct 2011 07:28:37 +0000 (09:28 +0200)]
tests: skip rather than failing when perl is not available
* tests/df/total-verify: Use require_perl_, so that this test is
skipped when perl is not available.
* tests/rm/deep-2: Likewise, and fix wording in a comment.
Reported by Bruno Haible.
Jim Meyering [Sat, 8 Oct 2011 17:57:21 +0000 (19:57 +0200)]
tests: avoid tail-2/F-vs-missing failure on NFS
* tests/tail-2/F-vs-missing: This test would fail about 80% of the
time on linux/powerpc when run in an NFS-mounted directory.
Reported by Bruno Haible in
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/1726/focus=1727
Jim Meyering [Sat, 8 Oct 2011 15:57:01 +0000 (17:57 +0200)]
tests: avoid ls/slink-acl test failure on Solaris 10
* tests/ls/slink-acl: Use setfacl's "-m user::r" option rather than
less-portable "-m m::r". The latter did not work with Solaris 10's
version of setfacl. Reported by Bruno Haible in
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/1726/focus=1737
Jim Meyering [Sat, 8 Oct 2011 15:11:10 +0000 (17:11 +0200)]
tests: avoid false-positive failures on linux/sparc64
* tests/init.cfg (require_strace_): Detect a version of strace
that fails on linux/sparc64 for 64-bit executables. Report and
most of the suggested fix from Bruno Haible:
http://thread.gmane.org/gmane.comp.gnu.coreutils.general/1726/focus=1728
build: let bootstrap resort to wget for downloading .po files
* bootstrap (download_po_files): Fallback to wget when downloading
the .po files via rsync failed. This is necessary to bootstrap behind
a strict firewall.
Jim Meyering [Mon, 3 Oct 2011 11:49:47 +0000 (13:49 +0200)]
tests: add a test to exercise today's ls-lL-vs-ACL bug
* tests/ls/slink-acl: New file.
* tests/Makefile.am (TESTS): Add it.
* tests/init.cfg (require_setfacl_): New function.
* gnulib: Update to latest, for file-has-acl changes.
* NEWS (Bug fixes): Mention it.
See http://thread.gmane.org/gmane.comp.lib.gnulib.bugs/28538. This
":>k; setfacl -m m::r k; ln -s k s; ls -Log s" should print e.g.,
-rw-r-----+ 1 0 Oct 5 19:22 s
With the ls from coreutils-8.13, it would print this (with "." or
nothing in place of the "+"):
-rw-r-----. 1 0 Oct 5 19:22 s
Markus Duft [Thu, 8 Sep 2011 11:09:27 +0000 (13:09 +0200)]
build: use getgr*_nomembers functions on Interix
Interix provides faster replacements for getgr{gid,nam,ent} where
group member information is not fetched from domain controllers.
This makes 'id' usable on domain controlled interix boxes.
* m4/jm-macros.m4: Check for _nomembers functions.
* src/system.h: Redefine function to _nomembers when available.
Markus Duft [Thu, 8 Sep 2011 09:18:17 +0000 (11:18 +0200)]
build: accommodate missing setgroups on Interix
Add a dummy, non-functional, always-successful replacement setgroups
function, to keep the original code untouched and simple.
* src/chroot.c (setgroups) [! HAVE_SETGROUPS]: Define.
Pádraig Brady [Sat, 1 Oct 2011 00:47:53 +0000 (01:47 +0100)]
ls: fix output of dangling symlinks colored as target
This is related to commit b7f2b51c, 2010-01-01,
"ls: fix color of broken symlinks colored as target"
which didn't handle the --dereference case.
The simplest way to reproduce the resultant
erroneous "argetm" is as follows:
$ ln -s /no-such dangle
$ env LS_COLORS=ln=target ls --dereference --color
ls: cannot access dangle: No such file or directory
argetmdangle
This is also an issue with the `tree` utility,
reported here: http://bugs.debian.org/586765
* src/ls.c (print_color_indicator): Move the handling
of 'ln=target' in $LS_COLORS (color_symlink_as_referent == true)
to a higher scope, to handle all cases where type == C_LINK.
* tests/misc/ls-misc: Add a test case for the specific issue,
and 2 further test cases to verify other code paths in this area.
Jim Meyering [Fri, 30 Sep 2011 18:13:01 +0000 (20:13 +0200)]
ls: avoid reading beyond end of color indicator
At first this looked like a buffer overrun, since there was no test
to ensure that the buffer length was 6. However, since the LS_COLORS
string is NUL-terminated and since settings within it are separated by
":" there was neither the risk of reading beyond end of buffer nor risk
of a false-positive match.
* src/ls.c (print_color_indicator): Use color_symlink_as_referent
rather than manually comparing against "target" again.
* src/system.h (STRNCMP_LIT): Correct description in comment.
Jim Meyering [Tue, 27 Sep 2011 14:32:35 +0000 (16:32 +0200)]
sort: avoid a NaN-induced infloop
These commands would fail to terminate:
yes -- -nan | head -156903 | sort -g > /dev/null
echo nan > F; sort -m -g F F
That can happen with any strtold implementation that includes
uninitialized data in its return value. The problem arises in the
mergefps function when bubble-sorting the two or more lines, each
from one of the input streams being merged: compare(a,b) returns 64,
yet compare(b,a) also returns a positive value. With a broken
comparison function like that, the bubble sort never terminates.
Why do the long-double bit strings corresponding to two identical
"nan" strings not compare equal? Because some parts of the result
are uninitialized and thus depend on the state of the stack.
For more details, see http://bugs.gnu.org/9612.
* src/sort.c (nan_compare): New function.
(general_numcompare): Use it rather than bare memcmp.
Reported by Aaron Denney in http://bugs.debian.org/642557.
* NEWS (Bug fixes): Mention it.
* tests/misc/sort-NaN-infloop: New file.
* tests/Makefile.am (TESTS): Add it.
Jim Meyering [Mon, 19 Sep 2011 18:13:36 +0000 (20:13 +0200)]
maint: avoid new syntax-check failure
Pulling in the latest gnulib triggered a new false-positive
syntax-check failure.
* cfg.mk (exclude_file_name_regexp--sc_prohibit_always-defined_macros):
Exempt remove.c; its definitions of DT_UNKNOWN, DT_DIR and DT_LNK are
harmless.
Bernhard Voelker [Mon, 19 Sep 2011 18:00:56 +0000 (20:00 +0200)]
tests: introduce make targets check-expensive and check-very-expensive
* Makefile.am: add shortcuts to run (very) expensive tests.
Use "make check-expensive" to run tests with RUN_EXPENSIVE_TESTS=yes,
use "make check-very-expensive" to run tests with both
RUN_EXPENSIVE_TESTS=yes and RUN_VERY_EXPENSIVE_TESTS=yes.
Non-expensive tests are included in all cases.
On some systems like glibc on GNU/kFreeBSD, a thread is
implicitly created when timer_settime() is used.
This breaks our scheme to ignore signals we've
sent ourselves.
* src/timeout.c (send_sig): Change the scheme used to
ignore signals we've sent ourselves, to a more robust
but perhaps limited scheme of ignoring all signals of
a certain type after we've sent that signal to the job.
* NEWS: Mention the change in behavior.
Jim Meyering [Mon, 19 Sep 2011 15:10:58 +0000 (17:10 +0200)]
gnulib: update for getcwd fix on systems without openat
This fixes a bug in pwd and all getcwd-using applications (for some
uses: df, readlink, stat) when run from a directory whose absolute name
contains more than PATH_MAX / 3 components. For more details, see
http://git.sv.gnu.org/cgit/gnulib.git/commit/?id=f6fe351fc534ae1
* gnulib: Update.
* NEWS (Improvements): Mention it.
* src/md5sum.c (split_3): Detect and handle BSD reversed
format checksums.
* tests/misc/md5sum-bsd: Add a new test.
* tests/Makefile.am: Reference new test.
* NEWS: Mention the improvement
Suggested by Rimas Kudelis.
Jim Meyering [Mon, 12 Sep 2011 13:04:32 +0000 (15:04 +0200)]
build: avoid unwarranted failure w/gcc-4.6.1 and --enable-gcc-warnings
* configure.ac (gl_GCC_VERSION_IFELSE): Define new macro.
(WERROR_CFLAGS): With --enable-gcc-warnings, use it to
add -Wsuggest-attribute=pure only with gcc 4.7 or newer.