]> Savannah Git Hosting - gnulib.git/log
gnulib.git
10 years agounistd: fix iOS check conditional
Daiki Ueno [Thu, 11 Dec 2014 07:07:31 +0000 (16:07 +0900)]
unistd: fix iOS check conditional

On Mac OS X 10.8, the TARGET_* macros are unconditonally defined
as 0 or 1 in <TargetConditionals.h>, and the previous check always
yielded true on non-iOS environment.
* lib/unistd.in.h (environ) [__APPLE__]: Check the values of
TARGET_OS_IPHONE and TARGET_IPHONE_SIMULATOR, instead of whether
they are defined.

10 years agoposixtm: avoid compiler warning in a better way
Paul Eggert [Wed, 10 Dec 2014 02:11:46 +0000 (18:11 -0800)]
posixtm: avoid compiler warning in a better way

* lib/posixtm.c (IF_LINT): Remove.
(year, posix_time_parse):
Return true (not 0) if successful.  All callers changed.
(posix_time_parse): Simplify to pacify GCC without need for IF_LINT.

10 years agorelocatable: support UNIXROOT in relocate() on EMX
KO Myung-Hun [Tue, 9 Dec 2014 01:40:48 +0000 (10:40 +0900)]
relocatable: support UNIXROOT in relocate() on EMX

UNIXROOT is used to specify a drive of a root of FHS. So if a path is
started with '/', then it should be translated to "$UNIXROOT/".

* lib/relocatable.c (relocate): Prepend $UNIXROOT to pathname if it is
started with '/' on EMX.

10 years agofreopen: workaround freopen() on OS/2 kLIBC
KO Myung-Hun [Fri, 5 Dec 2014 06:01:36 +0000 (15:01 +0900)]
freopen: workaround freopen() on OS/2 kLIBC

On OS/2 kLIBC, freopen() returns NULL even if it is successful if
filename is NULL.

* lib/freopen.c (rpl_freopen): Workaround.
* m4/freopen.m4: Add os2* case.

10 years agoget_shared_library_fullname: port to EMX
KO Myung-Hun [Wed, 3 Dec 2014 06:46:17 +0000 (15:46 +0900)]
get_shared_library_fullname: port to EMX

* lib/relocatable.c: Define strcmp and strncmp to stricmp and strnicmp
on EMX, respectively.
(_DLL_InitTerm): New on EMX.
(get_shared_library_fullname): Implement on EMX.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agofind_executable: port to EMX
KO Myung-Hun [Wed, 3 Dec 2014 07:02:50 +0000 (16:02 +0900)]
find_executable: port to EMX

* lib/progreloc.c (find_executable): Implement on EMX.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agosched: check struct sched_param in spawn.h as well
KO Myung-Hun [Thu, 4 Dec 2014 01:03:34 +0000 (10:03 +0900)]
sched: check struct sched_param in spawn.h as well

On OS/2 kLIBC, struct sched_param is in spawn.h. So without this a
redefinition error occurs when sched.h is inclued by spawn.h.

* lib/sched.in.h: Include spawn.h on kLIBC.
* lib/sched_h.m4: Check struct sched_param in spawn.h as well.

10 years agosupport GNU format printf and scanf on mingw
Pádraig Brady [Tue, 2 Dec 2014 13:34:06 +0000 (13:34 +0000)]
support GNU format printf and scanf on mingw

Allow %lld and %Lf to work on mingw as per:
http://sourceforge.net/p/mingw-w64/wiki2/gnu%20printf/

10 years agobootstrap: Allow perl modules in $buildreq
Martin Kletzander [Mon, 8 Dec 2014 12:19:12 +0000 (13:19 +0100)]
bootstrap: Allow perl modules in $buildreq

With this patch it is possible to put e.g. "perl::XML:XPath -" in
$buildreq in bootstrap.conf which will cause a check for perl module
XML::XPath using:

  perl -m"XML::XPath" -e 'exit 0' >/dev/null 2>&1

If this fails due to any other error than XML::XPath missing, it is left
on the user to fix up his/her bootstrap.conf.  One of the examples
might be perl itself missing, which should be in $buildreq and precede
any perl::Module specifications.  Versioning of perl modules is not
supported.

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
10 years agoapply _GL_ATTRIBUTE_PURE to some inline functions
Pádraig Brady [Sat, 6 Dec 2014 01:28:17 +0000 (01:28 +0000)]
apply _GL_ATTRIBUTE_PURE to some inline functions

clang 3.4.2 flagged these inline functions as pure

* lib/savewd.h (savewd_errno): Set _GL_ATTRIBUTE_PURE.
* lib/sig-handler.h (get_handler): Likewise.
* lib/stat-time.h (get_stat_{a,c,m,birth}time{,_ns}): Likewise.
* lib/timespec.h (timespec_cmp, timespec_sign): Likewise.

10 years agovasnprintf: fix potential incorrect errno
Pádraig Brady [Mon, 8 Dec 2014 09:46:38 +0000 (09:46 +0000)]
vasnprintf: fix potential incorrect errno

An adjustment of the previous commit c5c4f53b.

* lib/vasnprintf.c (VASNPRINTF): free() generally doesn't set errno,
but it can potentially in certain edge cases.
Reported by Eric Blake.

10 years agovasnprintf: fix potential use after free
Pádraig Brady [Sat, 6 Dec 2014 01:14:02 +0000 (01:14 +0000)]
vasnprintf: fix potential use after free

* lib/vasnprintf.c (VASNPRINTF): Fix free-memory read,
flagged by clang-analyzer 3.4.2.

10 years agoautoupdate
Karl Berry [Fri, 5 Dec 2014 13:56:18 +0000 (05:56 -0800)]
autoupdate

10 years agofilevercmp, posixtm: avoid compiler warnings with -O3
Pádraig Brady [Fri, 5 Dec 2014 12:32:49 +0000 (12:32 +0000)]
filevercmp, posixtm: avoid compiler warnings with -O3

* lib/filevercmp.h (filevercmp): Tag with _GL_ATTRIBUTE_PURE
* lib/posixtm.c: (IF_LINT): Define.
(posix_time_parse): Use it to void a "may be used uninitialized"
warning, seen only with -O3.

10 years agoFix LDBL80_WORDS macro on big endian platforms.
Bruno Haible [Fri, 5 Dec 2014 10:30:27 +0000 (11:30 +0100)]
Fix LDBL80_WORDS macro on big endian platforms.

* m4/isfinite.m4 (gl_ISFINITEL_WORKS): Add missing parenthesis in
LDBL80_WORDS macro.
* m4/isinf.m4 (gl_ISINFL_WORKS): Likewise.
* m4/isnanl.m4 (gl_FUNC_ISNANL_WORKS): Likewise.
* tests/test-isfinite.c (test_isfinitel): Likewise.
* tests/test-isinf.c (test_isinfl): Likewise.
* tests/test-isnan.c (test_long_double): Likewise.
* tests/test-isnanl.h (main): Likewise.
* tests/test-snprintf-posix.h (LDBL80_WORDS): Add missing parenthesis.
* tests/test-sprintf-posix.h (LDBL80_WORDS): Likewise.
* tests/test-vasnprintf-posix.c (LDBL80_WORDS): Likewise.
* tests/test-vasprintf-posix.c (LDBL80_WORDS): Likewise.
Reported by Pádraig Brady.

10 years agoautoupdate
Karl Berry [Wed, 3 Dec 2014 15:04:40 +0000 (07:04 -0800)]
autoupdate

10 years agogit-version-gen: do not print new line characters
KO Myung-Hun [Tue, 2 Dec 2014 05:55:20 +0000 (14:55 +0900)]
git-version-gen: do not print new line characters

On platforms with CRLF endings (such as OS/2), use of 'echo' may
leave behind an unwanted CR.

* build-aux/git-version-gen: Use printf instead of echo and tr.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agognulib-tool: recognize x:* as an absolute path
KO Myung-Hun [Tue, 2 Dec 2014 03:50:49 +0000 (12:50 +0900)]
gnulib-tool: recognize x:* as an absolute path

On OS/2, x:* is an absolute path, too.

* gnulib-tool (func_gnulib_dir): Add ?:* case.
(func_relconcat): Likewise.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agoargp: avoid extraneous translation and mem leak with empty pre doc
Andrei Borzenkov [Tue, 2 Dec 2014 16:05:10 +0000 (16:05 +0000)]
argp: avoid extraneous translation and mem leak with empty pre doc

* lib/argp-help.c (argp_doc): Never translate the empty string,
when "\v" is the first or last character of the string, as that
has a reserved meaning to return the header info from a po file.
This also fixes a small memory leak in the !post case.
The issue can be seen with this command for example:
LC_MESSAGES=en_US grub2-mknetdir --help

10 years agoautoupdate
Karl Berry [Sat, 29 Nov 2014 14:00:25 +0000 (06:00 -0800)]
autoupdate

10 years agodoc: mention that _BSD_SOURCE is deprecated for _DEFAULT_SOURCE
Pádraig Brady [Fri, 28 Nov 2014 10:57:31 +0000 (10:57 +0000)]
doc: mention that _BSD_SOURCE is deprecated for _DEFAULT_SOURCE

* doc/posix-functions/setjmp.texi: Defining _BSD_SOURCE will induce
warnings unless _DEFAULT_SOURCE is also defined.

10 years agouniname/uniname-tests: skip if system's libunistring is used
Daiki Ueno [Thu, 27 Nov 2014 09:46:43 +0000 (18:46 +0900)]
uniname/uniname-tests: skip if system's libunistring is used

* modules/uniname/uniname-tests (Makefile.am): Skip test if
uniname/uniname module is not compiled.

10 years agoprintf: fix configure check on big endian systems
Pádraig Brady [Thu, 27 Nov 2014 11:17:07 +0000 (11:17 +0000)]
printf: fix configure check on big endian systems

* m4/printf.m4 (gl_PRINTF_INFINITE_LONG_DOUBLE): Add missing bracket.

10 years agopipe-filter-gi, pipe-filter-ii: port to AIX
Daiki Ueno [Tue, 18 Nov 2014 12:11:09 +0000 (21:11 +0900)]
pipe-filter-gi, pipe-filter-ii: port to AIX

On AIX 7.1, 'select' is defined as static and cannot be referred
to from inline function.
* lib/pipe-filter-aux.h (nointr_select): Remove, manually expand
the definition...
* lib/pipe-filter-gi.c (filter_loop): ...here, and...
* lib/pipe-filter-ii.c (pipe_filter_ii_execute): ...here.

10 years agogitlog-to-changelog: add --until
Paul Eggert [Thu, 20 Nov 2014 17:29:35 +0000 (09:29 -0800)]
gitlog-to-changelog: add --until

* build-aux/gitlog-to-changelog: Support new --until option.
Need described by Eli Zaretskii in: http://bugs.gnu.org/19113

10 years agoupdate from texinfo
Karl Berry [Sun, 16 Nov 2014 18:41:11 +0000 (10:41 -0800)]
update from texinfo

10 years agoextern-inline: update commentary about GCC bugs
Paul Eggert [Fri, 14 Nov 2014 20:10:50 +0000 (12:10 -0800)]
extern-inline: update commentary about GCC bugs

* m4/extern-inline.m4: Add another GCC bug number to comments.

10 years agogen-uni-tables: untabify
Daiki Ueno [Thu, 13 Nov 2014 07:54:53 +0000 (16:54 +0900)]
gen-uni-tables: untabify

* lib/gen-uni-tables.c: Untabify.

10 years agogen-uni-tables: check out-of-range values added to 3-level tables
Daiki Ueno [Fri, 10 Oct 2014 13:59:46 +0000 (22:59 +0900)]
gen-uni-tables: check out-of-range values added to 3-level tables

* lib/gen-uni-tables.c (output_category, output_bidi_category)
(output_joining_type, output_ident_category): Check out-of-range
values added to 3-level tables.

10 years agogen-uni-tables: utilize 'assert'
Daiki Ueno [Thu, 13 Nov 2014 06:56:01 +0000 (15:56 +0900)]
gen-uni-tables: utilize 'assert'

* lib/gen-uni-tables.c: Include <assert.h>.
(output_category, output_combclass, output_decimal_digit_test)
(output_decimal_digit, output_digit_test, output_digit)
(output_numeric, get_mirror_value, fill_properties)
(fill_property30, is_property_alphabetic)
(is_property_default_ignorable_code_point)
(is_property_uppercase, is_property_lowercase)
(is_property_cased, is_property_case_ignorable)
(is_property_changes_when_lowercased, is_property_iso_control)
(is_property_math, fill_arabicshaping, output_joining_group)
(fill_scripts, fill_blocks, output_lbp, fill_org_wbp)
(output_wbp, fill_org_gbp, get_decomposition)
(output_decomposition, fill_composition_exclusions)
(debug_output_composition_tables, output_composition_tables)
(redistribute_casefolding_rules, output_casing_rules): Use
'assert (EXPR);' instead of 'if (!EXPR) abort ();' for better error
reporting.

10 years agogen-uni-tables: cosmetic improvements
Daiki Ueno [Thu, 23 Oct 2014 07:05:37 +0000 (16:05 +0900)]
gen-uni-tables: cosmetic improvements

* lib/gen-uni-tables.c: Escape newlines in the Emacs file local
variables specification.
(is_outdigit): Remove unused function.

10 years agofcntl-h-tests: port to PA-RISC GNU/Linux
Jeroen Roovers [Thu, 13 Nov 2014 02:55:49 +0000 (18:55 -0800)]
fcntl-h-tests: port to PA-RISC GNU/Linux

* tests/test-fcntl-h.c (main): Check O_SYNC is different from O_DSYNC.

10 years agofts: port to C89
Paul Eggert [Mon, 10 Nov 2014 23:43:22 +0000 (15:43 -0800)]
fts: port to C89

Problem reported for MSVC 16 by Gisle Vanem in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-11/msg00027.html
* lib/fts.c (fts_build): Avoid declaration before statement.

10 years agounistd: port to iOS
Paul Eggert [Fri, 7 Nov 2014 04:33:02 +0000 (20:33 -0800)]
unistd: port to iOS

Problem reported by André Klitzing in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-11/msg00013.html
* lib/unistd.in.h (environ): Do not include crt_externs.h on iOS.

10 years agoobstack: do not reject malloc-style obstack_chunkfun, obstack_freefun
Paul Eggert [Wed, 5 Nov 2014 19:41:57 +0000 (11:41 -0800)]
obstack: do not reject malloc-style obstack_chunkfun, obstack_freefun

Problem reported by Alan Modra in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-11/msg00007.html
* lib/obstack.h (obstack_chunkfun, obstack_freefun):
Always cast the function arg, reverting this part of the previous
change.

10 years agoautoupdate
Karl Berry [Wed, 5 Nov 2014 14:01:18 +0000 (06:01 -0800)]
autoupdate

10 years agoupdate from texinfo
Karl Berry [Tue, 4 Nov 2014 18:01:31 +0000 (10:01 -0800)]
update from texinfo

10 years agoobstack: avoid potentially-nonportable function casts
Paul Eggert [Tue, 4 Nov 2014 08:31:31 +0000 (00:31 -0800)]
obstack: avoid potentially-nonportable function casts

* lib/obstack.c (CALL_CHUNKFUN, CALL_FREEFUN):
Remove, replacing with ...
(call_chunkfun, call_freefun): New static functions.
All uses changed.  Avoid potentially-nonportable casts.
(chunkfun_type, freefun_type): Remove typedefs; no longer used.
(_obstack_begin_worker): Omit last two args, since they
rely on potentially-nonportable casts.  All callers changed.
* lib/obstack.h (_OBSTACK_CAST): New macro.
Use it everywhere the old API used a potentially-nonportable cast.
The new API doesn't cast.
(struct obstack): Use unions rather than requiring
potentially-nonportable casts.
(obstack_chunkfun, obstack_freefun): Return void.

10 years agoobstack: fix macro return values
Alan Modra [Tue, 4 Nov 2014 01:32:27 +0000 (17:32 -0800)]
obstack: fix macro return values

* lib/obstack.h (obstack_next_free): Return void *.
(obstack_1grow_fast, obstack_blank_fast): Return void.
For __GNUC__ macros:
(obstack_1grow, obstack_blank): Remove now unnecessary (void) 0.
For !__GNUC__ macros:
(obstack_make_room, obstack_grow, obstack_grow0)
(obstack_ptr_grow_fast, obstack_int_grow_fast): Return void.

10 years agoobstack: do not assume system-supplied obstack is size_t safe
Paul Eggert [Tue, 4 Nov 2014 00:36:27 +0000 (16:36 -0800)]
obstack: do not assume system-supplied obstack is size_t safe

* m4/obstack.m4: New file.
* modules/obstack (Files): Add it.

10 years agoobstack: port to platforms that #define __alignof__
Paul Eggert [Tue, 4 Nov 2014 00:34:52 +0000 (16:34 -0800)]
obstack: port to platforms that #define __alignof__

* lib/obstack.c: Include <alignof.h> if !defined __alignof__,
not if !_LIBC.  We don't know of any platforms that #define
__alignof__, but it might be useful in tests.  Conversely,
glibc assumes GCC.

10 years agolinkat: don't unconditionally replace on GNU/Linux
Pádraig Brady [Fri, 31 Oct 2014 18:18:36 +0000 (18:18 +0000)]
linkat: don't unconditionally replace on GNU/Linux

* m4/linkat.m4 (gl_FUNC_LINKAT): The compile check for AT_SYMLINK_FOLLOW
was redundant for a few reasons.  It was present to support compiling
on new systems but running on the old narrow window of Linux 2.6.1[67].
It setup and cleaned up test files which weren't actually used.
On non __linux__ it compile tested AT_SYMLINK_FOLLOW, but that is
implicit in the following check.

10 years agolinkat: wrap to handle symlinks on OS X 10.10
Pádraig Brady [Fri, 31 Oct 2014 18:12:15 +0000 (18:12 +0000)]
linkat: wrap to handle symlinks on OS X 10.10

* m4/linkat.m4 (gl_FUNC_LINKAT): linkat() is available on Yosemite
but not usable because it doesn't support creating hardlinks
to symlinks.  Therefore add a generic test for this capability
and fallback to our emulation if linkat() fails with ENOTSUP.

10 years agoopen, openat: document nonstandard FreeBSD, NetBSD O_NOFOLLOW errno
Paul Eggert [Mon, 3 Nov 2014 06:24:09 +0000 (22:24 -0800)]
open, openat: document nonstandard FreeBSD, NetBSD O_NOFOLLOW errno

* doc/posix-functions/open.texi (open):
* doc/posix-functions/openat.texi (openat):
Document that these functions do not set errno to ELOOP when
a symlink is opened with O_NOFOLLOW.

10 years agoobstack: add NEWS entry for recent incompatible changes
Paul Eggert [Fri, 31 Oct 2014 19:45:22 +0000 (12:45 -0700)]
obstack: add NEWS entry for recent incompatible changes

* NEWS: Describe recent changes.

10 years agomountlist: don't use libmount to decide on dummy/remote
Pádraig Brady [Thu, 30 Oct 2014 04:08:50 +0000 (04:08 +0000)]
mountlist: don't use libmount to decide on dummy/remote

* lib/mountlist.c (read_file_system_list): Don't use the libmount
routines to determine whether a file system is dummy or remote,
as they're not currently compatible.  For example the remoteness
is determined on file system type (for which the list seems incomplete),
rather than simply checking for a ':' in the device name.
Also libmount currently determines that 'tmpfs' is a dummy file system
even though it has associated storage.

10 years agomaint: add missing ChangeLog entries for Modra's obstack changes
Paul Eggert [Thu, 30 Oct 2014 01:03:53 +0000 (18:03 -0700)]
maint: add missing ChangeLog entries for Modra's obstack changes

See: http://lists.gnu.org/archive/html/bug-gnulib/2014-10/msg00072.html

10 years agoobstack: prefer __alignof__ to alignof
Paul Eggert [Wed, 29 Oct 2014 23:15:41 +0000 (16:15 -0700)]
obstack: prefer __alignof__ to alignof

This is for portability to pre-4.7 GCC when compiling glibc.
See Joseph S. Myers in:
http://sourceware.org/ml/libc-alpha/2014-10/msg00703.html
* lib/obstack.c (__alignof__) [!_LIBC && !__GNUC__]:
New macro, defined by including and using <alignof.h>.
(MAX): New macro.
(DEFAULT_ALIGNMENT, DEFAULT_ROUNDING): Redefine in terms of these.
Do not use enums as they are not portable to some broken compilers.
* modules/obstack (Depends-on): Depend on alignof, not stdalign.

10 years agoobstack: prefer alignof to calculating alignments by hand
Paul Eggert [Wed, 29 Oct 2014 07:22:08 +0000 (00:22 -0700)]
obstack: prefer alignof to calculating alignments by hand

* lib/obstack.c: Include <stdalign.h>.
(struct fooalign): Remove.
(DEFAULT_ALIGNMENT): Use alignof rather than the old offsetof hack.
* modules/obstack (Depends-on): Add stdalign.

10 years agoobstack: use size_t alignments and check for overflow
Paul Eggert [Wed, 29 Oct 2014 06:58:42 +0000 (23:58 -0700)]
obstack: use size_t alignments and check for overflow

* lib/obstack.c, lib/obstack.h (_obstack_begin, _obstack_begin_1):
* lib/obstack.c (_obstack_begin_worker, _obstack_newchunk):
* lib/obstack.h (struct obstack.alignment_mask):
Use _OBSTACK_SIZE_T, not int, for alignments.
* lib/obstack.c (_obstack_newchunk): Fail if the size calculation
overflows, e.g., when adding the alignment.

10 years agoobstack: 64-bit obstack support, part 3
Alan Modra [Wed, 29 Oct 2014 03:33:00 +0000 (14:03 +1030)]
obstack: 64-bit obstack support, part 3

This finally enables full 64-bit obstack support.  The glibc shared library
specific code is removed from obstack.c too, and the error handling code
conditionally compiled under control of another macro,
_OBSTACK_NO_ERROR_HANDLER.

* lib/obstack.h: Include string.h earlier.
(_OBSTACK_INTERFACE_VERSION): Define.
(_OBSTACK_SIZE_T, _CHUNK_SIZE_T): Define as size_t for version 2.
* lib/obstack.c: Don't include shlib-compat.h.
(OBSTACK_INTERFACE_VERSION): Delete.
(_OBSTACK_ELIDE_CODE): Rename from ELIDE_CODE.  Define when version 1
glibc code is compatible with version 2.  Don't include stdio.h for
__GNU_LIBRARY.
(obstack_exit_failure, print_and_abort, obstack_alloc_failed_handler):
Omit when _OBSTACK_NO_ERROR_HANDLER defined.  Include stdio.h here.
(_obstack_compat, _obstack, _obstack_newchunk, obstack_free): Delete
glibc shared library specific source.

10 years agoobstack: 64-bit obstack support, part 2
Alan Modra [Wed, 29 Oct 2014 03:32:40 +0000 (14:02 +1030)]
obstack: 64-bit obstack support, part 2

This gets us 4G obstack support, without changing ABI compatibility,
apart from possibly introducing some signed/unsigned comparison
warnings in code that uses obstack.h.

a) Replace "int" size parameters, return values, and macro local vars
   with _OBSTACK_SIZE_T, an "unsigned int" for now.
b) Make obstack.chunk_size a _CHUNK_SIZE_T, an "unsigned long" for now.
c) Make all obstack macros checking available room use obstack_room.
   "next_free + desired > chunk_limit" may wrap the lhs for chunks
   allocated near the top of memory.
d) Use unsigned comparisons, and macro locals to support >2G on 32-bit.

* lib/obstack.h (_OBSTACK_SIZE_T): Define.  Use throughout
in place of "int" size parameters, return values and local vars.
(_CHUNK_SIZE_T): Define.
(struct obstack): Make chunk_size a _CHUNK_SIZE_T.  Make temp
union use an _OBSTACK_SIZE_T integer type.
For __GNUC__ versions of the following macros...
(obstack_room): Rename local var.
(obstack_make_room): Use obstack_room.
(obstack_grow, obstack_grow0, obstack_1grow, obstack_ptr_grow,
obstack_int_grow, obstack_blank): Likewise.
(obstack_finish): Use unsigned comparison when comparing aligned
next_free against chunk_limit.
(obstack_free): Cast OBJ to remove possible const qualifier.
For !__GNUC__ versions of the following macros...
(obstack_make_room): Use obstack_room.
(obstack_grow, obstack_grow0, obstack_1grow, obstack_ptr_grow,
obstack_int_grow, obstack_blank): Likewise.
(obstack_finish): Use unsigned comparision when comparing aligned
next_free against chunk_limit.
(obstack_free): Use temp.p and same comparisons as __GNUC__ version.
* lib/obstack.c (_obstack_begin_worker): Make "size" parameter
_OBSTACK_SIZE_T.
(_obstack_begin, _obstack_begin_1): Likewise.
(_obstack_newchunk): Likewise for length parameter.  Use size_t locals.
(_obstack_memory_used): Return and use _OBSTACK_SIZE_T local.

10 years agoobstack: 64-bit obstack support, part 1
Alan Modra [Wed, 29 Oct 2014 03:32:31 +0000 (14:02 +1030)]
obstack: 64-bit obstack support, part 1

a) Correct calls to alloc function, to use a size_t arg.  "long" is
   just wrong on targets like x86_64-mingw64 where "long" is 32 bits
   and "size_t" 64 bits.
b) Consolidate _obstack_begin and _obstack_begin1 code.

* lib/obstack.h (struct obstack <chunkfun>): Correct prototype to
use "size_t" rather than "long".
(_obstack_begin, _obstack_begin1): Likewise.
(obstack_init, obstack_begin, obstack_specify_allocation_with_arg,
obstack_chunkfun): Update alloc function casts.
* lib/obstack.c (CALL_CHUNKFUN): Update chunkfun cast.
(chunkfun_type, freefun_type): New typdefs.
(_obstack_begin_worker): Split out from ..
(_obstack_begin, _obstack_begin_1): ..here.

10 years agoobstack: tidy part 2
Alan Modra [Wed, 29 Oct 2014 03:32:22 +0000 (14:02 +1030)]
obstack: tidy part 2

a) Don't be concerned about "not polluting the namespace with stddef.h
   symbols" in obstack.h, since gnulib string.h includes stddef.h
   anyway, and it seems unlikely that anyone would care.
b) Don't roll our own slow memcpy in _obstack_newchunk.
c) Rename obstack_free to _obstack_free.  This makes the naming
   consistent with other obstack functions and obviates the need for
   __obstack_free.  Ancient obstack.c defined both obstack_free and
   _obstack_free.  We continue to do that for _LIBC via an alias.
d) Miscellaneous macro fixes.  The expression used to test for gcc-2.8
   is clever, but nowadays gcc warns on undefined macros.  You'll get
   an undefined macro warning if simulating an old gcc with -U__GNUC__
   -U__GNUC_MINOR__ -D__GNUC__=1.

* lib/obstack.h: Include stddef.h unconditionally.  Formatting fixes.
(PTR_INT_TYPE): Delete, replace with ptrdiff_t.
(__obstack_free): Delete, update refs.
(_obstack_free): Rename from obstack_free.
(__extension__): Avoid undefined macro warning for __GNUC_MINOR__.
(obstack_object_size, obstack_room): Parenthesise !__GNUC__ versions.
* lib/obstack.c: Don't include stddef.h.
(COPYING_UNIT): Delete.
(_obstack_begin): Formatting fix.
(_obstack_newchunk): Use memcpy to move existing object to new chunk.
(_obstack_free): Rename from __obstack_free, update alias.  Move
undef of obstack_free to where it is needed.

10 years agoobstack: tidy part 1
Alan Modra [Wed, 29 Oct 2014 03:32:13 +0000 (14:02 +1030)]
obstack: tidy part 1

a) Rename temp fields.  temp.tempint and temp.tempptr just looks ugly
   to me, and result in overlong lines after later patches.
b) Move error handling code, to avoid a forward declaration and to
   simplify later patches in this series.

* lib/obstack.h (struct obstack <temp>): Rename fields of union
and update all uses.
* lib/obstack.c: Include stdlib.h earlier.
(obstack_exit_failure, obstack_alloc_failed_handler): Move later
in file.
(print_and_abort): Remove now redundant forward declaration.

10 years agosocketlib, sockets, sys_socket: Use AC_REQUIRE to pacify autoconf.
Paul Eggert [Fri, 24 Oct 2014 17:47:44 +0000 (10:47 -0700)]
socketlib, sockets, sys_socket: Use AC_REQUIRE to pacify autoconf.

Without this change, in bleeding-edge fileutils Autoconf complains
that gl_SOCKETLIB etc. are AC_REQUIREd after being invoked.
* modules/socketlib (configure.ac): AC_REQUIRE gl_SOCKETLIB.
* modules/sockets (configure.ac): AC_REQUIRE gl_SOCKETS.
* modules/sys_socket (configure.ac): AC_REQUIRE gl_HEADER_SYS_SOCKET.

10 years agoiconv: avoid false detection of non-working iconv
Daiki Ueno [Tue, 21 Oct 2014 09:07:00 +0000 (18:07 +0900)]
iconv: avoid false detection of non-working iconv

The INBUF arguments of iconv can be either 'const char **'
or 'char **'.  If CC is g++, the difference causes a compile error
and thus leads to a false detection of non-working iconv.
Reported by Eli Zaretskii and Werner LEMBERG in:
<https://lists.gnu.org/archive/html/bug-gnulib/2014-10/msg00023.html>.
* m4/iconv.m4 (AM_ICONV_LINK): Try all possible argument types of
iconv.  Bump serial number.

10 years agobootstrap: print more diagnostics for missing programs
Pádraig Brady [Thu, 23 Oct 2014 12:59:08 +0000 (13:59 +0100)]
bootstrap: print more diagnostics for missing programs

* build-aux/bootstrap: only suppress stderr when checking for
alternative program names.  This supports programs issuing non
standard error messages like:
  "Provide an AUTOMAKE_VERSION environment variable, please"
Reported by Ingo Schwarze with OpenBSD

10 years agobootstrap: only update the gnulib submodule
Pádraig Brady [Thu, 23 Oct 2014 12:26:23 +0000 (13:26 +0100)]
bootstrap: only update the gnulib submodule

* build-aux/bootstrap: Restrict the "submodule update" command
to the gnulib path.

Reported by Reuben Thomas

10 years agosymlinkat: port to AIX 7.1
Paul Eggert [Sun, 19 Oct 2014 06:22:36 +0000 (23:22 -0700)]
symlinkat: port to AIX 7.1

* doc/posix-functions/symlinkat.texi (symlinkat):
Mention AIX porting problem.
* lib/symlinkat.c: Always include errno.h.
(rpl_symlinkat) [HAVE_SYMLINKAT]: New function.
* lib/unistd.in.h (symlinkat): Add replacement machinery.
* m4/symlinkat.m4 (gl_FUNC_SYMLINKAT): Check symlinkat behavior.
* m4/unistd_h.m4 (gl_UNISTD_H_DEFAULTS): Add REPLACE_SYMLINKAT.
* modules/symlinkat (Depends-on): Add fstatat if REPLACE_SYMLINKAT.
(configure.ac): Also compile replacement if REPLACE_SYMLINKAT.
* modules/unistd (unistd.h): Substitute REPLACE_SYMLINKAT.

10 years agoreadlinkat: port to AIX 7.1
Paul Eggert [Sun, 19 Oct 2014 05:35:28 +0000 (22:35 -0700)]
readlinkat: port to AIX 7.1

* doc/posix-functions/readlink.texi (readlink):
* doc/posix-functions/readlinkat.texi (readlinkat):
Mention AIX porting problem.
* lib/readlinkat.c (rpl_readlinkat) [HAVE_READLINKAT]:
New function.
* lib/unistd.in.h (readlinkat): Add replacement machinery.
* m4/readlinkat.m4 (gl_FUNC_READLINKAT): Check readlinkat signature.
* m4/unistd_h.m4 (gl_UNISTD_H_DEFAULTS): Add REPLACE_READLINKAT.
* modules/readlinkat (configure.ac): Also compile replacement
if REPLACE_READLINKAT.
* modules/unistd (unistd.h): Substitute REPLACE_READLINKAT.

10 years agoremove spurious {
Karl Berry [Sun, 12 Oct 2014 16:03:23 +0000 (09:03 -0700)]
remove spurious {

10 years agomodules/fcntl: fix error reporting by dupfd
Eli Zaretskii [Tue, 7 Oct 2014 17:59:11 +0000 (20:59 +0300)]
modules/fcntl: fix error reporting by dupfd

* lib/fcntl.c (dupfd) [_WIN32]: Don't overwrite the value of
errno set by _open_osfhandle by EMFILE.  Convert errors
returned by DuplicateHandle to corresponding errno values.

10 years agobasename, dirname: Improve documentation.
Ben Pfaff [Sun, 28 Sep 2014 16:52:08 +0000 (09:52 -0700)]
basename, dirname: Improve documentation.

* doc/posix-functions/basename.texi: Mention dirname module and
base_name() function.
* doc/posix-functions/dirname.texi: Mention dir_name() and
mdir_name() functions.

Suggested by Werner LEMBERG <wl@gnu.org>.

10 years agoexclude: declare exclude_patopts static
Jim Meyering [Wed, 24 Sep 2014 16:03:06 +0000 (09:03 -0700)]
exclude: declare exclude_patopts static

* lib/exclude.c (exclude_patopts): Declare static,
to avoid triggering a -Wmissing-prototypes warning.
The alternative (declaring it in the .h file) would
require publicizing the private "struct patopts".

10 years agoautoupdate
Karl Berry [Fri, 26 Sep 2014 16:07:05 +0000 (09:07 -0700)]
autoupdate

10 years agodirname: support compilation with C++
Werner LEMBERG [Sun, 21 Sep 2014 15:10:15 +0000 (17:10 +0200)]
dirname: support compilation with C++

* lib/dirname.h: Add necessary C linkage declarations.

10 years agoqsort_r: include <config.h>
Paul Eggert [Tue, 16 Sep 2014 19:26:48 +0000 (12:26 -0700)]
qsort_r: include <config.h>

Problem reported by Tom G. Christensen in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-09/msg00071.html
* lib/qsort.c [!_LIBC]: Include <config.h> first.

10 years agoavltree-list: avoid compiler warnings
Dylan Cali [Tue, 16 Sep 2014 15:42:36 +0000 (16:42 +0100)]
avltree-list: avoid compiler warnings

* lib/gl_anytree_list2.h: Add _GL_ATTRIBUTE_PURE to avoid
-Werror=suggest-attribute=pure.
* lib/gl_array_list.c: Likewise.
* lib/gl_avltree_list.c (gl_avltree_list_check_invariants): Add extern
declaration to avoid -Werror=missing-prototypes.  This is not added
to a header as only exported for tests.  Add (void) to the
check_invariants() call to indicate we're discarding the result
in this context which avoids -Werror=unused-value.  Note we don't
use ignore_value here to avoid a dependency as we know we'll not
be adding __attribute__((warn_unused_result)) to check_invariants().
Add _GL_ATTRIBUTE_CONST to avoid -Werror=suggest-attribute=const.

10 years agoqsort_r: new module, for GNU-style qsort_r
Paul Eggert [Mon, 15 Sep 2014 15:59:18 +0000 (08:59 -0700)]
qsort_r: new module, for GNU-style qsort_r

* m4/qsort_r.m4: New file.  Forgot to add this earlier.

10 years agostrerror_r-posix: support compilation with C++
Werner LEMBERG [Mon, 15 Sep 2014 12:41:57 +0000 (13:41 +0100)]
strerror_r-posix: support compilation with C++

* lib/strerror_r.c: Add necessary C linkage declarations.

10 years agofcntl-h: fix compilation with Intel C++ compiler
Johannes Zarl [Thu, 11 Sep 2014 13:38:53 +0000 (14:38 +0100)]
fcntl-h: fix compilation with Intel C++ compiler

* lib/fcntl.in.h: ICC has the same issue as GCC <= 4.2.

10 years agoautoupdate
Karl Berry [Thu, 11 Sep 2014 13:01:12 +0000 (06:01 -0700)]
autoupdate

10 years agomountlist: use /proc/self/mountinfo when available
Fridolin Pokorny [Wed, 27 Aug 2014 13:25:30 +0000 (15:25 +0200)]
mountlist: use /proc/self/mountinfo when available

Use libmount to propagate device IDs provided by Linux in
/proc/self/mountinfo.  This will give more accurate output when
using df in chroot'ed environments as the device IDs are not
determined by stat() which may be inaccurate within the chroot.

* lib/mountlist.c (read_file_system_list): Use the libmount routines
from util-linux to parse "/proc/self/mountinfo" or fall back to
standard getmntent() processing.
* m4/ls-mntd-fs.m4: Check for libmount only when 1-argument
getmntent() is used, as is the case on GNU/Linux.
* DEPENDENCIES: Mention the optional util-linux dependency.

10 years agousers.txt: add cmogstored
Eric Wong [Sun, 7 Sep 2014 03:22:12 +0000 (03:22 +0000)]
users.txt: add cmogstored

cmogstored has used gnulib since the beginning in 2012 to support
GNU/Linux, FreeBSD, and GNU/kFreeBSD.  cmogstored is currently
included in the FreeBSD ports collection, but developed primarily
on GNU/Linux.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
10 years agognulib-tool: Sync with build-aux/bootstrap options
Mathieu Anquetin [Thu, 4 Sep 2014 06:54:49 +0000 (08:54 +0200)]
gnulib-tool: Sync with build-aux/bootstrap options

Options for downloading PO files were not in sync with their
build-aux/bootstrap counterpart. This commit adopt the same behavior for
these tools.

Signed-off-by: Mathieu Anquetin <mathieu@anquetin.eu>
10 years agognulib-tool: Fallback to wget when rsync fails
Mathieu Anquetin [Thu, 4 Sep 2014 06:54:48 +0000 (08:54 +0200)]
gnulib-tool: Fallback to wget when rsync fails

Current implementation only tries to rsync PO files when rsync is
installed on the host. In case of error, no files are downloaded even
if they are available. This leads to bootstrap problems for hosts
that lie behind a restrictive firewall.

This patch always tries to rsync by default, falling back to wget if
an error occurs.

Signed-off-by: Mathieu Anquetin <mathieu@anquetin.eu>
10 years agomaintainer-makefile: add syntax check for useless ';;'
Eric Blake [Thu, 4 Sep 2014 19:11:04 +0000 (13:11 -0600)]
maintainer-makefile: add syntax check for useless ';;'

Most instances of ;; in C code are mistakes, where the second
semicolon is a no-op.  This rule tries to make it easy to flag
the typos occuring at the end of a statement.  It intentionally
does not flag for(;;) loops, and misses grammar problems in
comments if the problem occurs in the middle of the line.

Shell files (including configure.ac and Makefile.am, which can
contain shell snippets) are too likely to use case statements
where ;; is legitimate, so those are not scanned.

* top/maint.mk (sc_prohibit_double_semicolon): New rule.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agopthread, pthread_sigmask, threadlib: port to Ubuntu 14.04
Paul Eggert [Thu, 4 Sep 2014 21:55:12 +0000 (14:55 -0700)]
pthread, pthread_sigmask, threadlib: port to Ubuntu 14.04

Problem reported by Assaf Gordon in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-09/msg00023.html
Apparently Ubuntu is doing some fancy link-time optimization
that doesn't work with -lpthread but does work with -pthread.
Work around the bug by preferring -pthread to -lpthread.
This change affects only LIBS, not CFLAGS, which is a little
weird, but it works.
* m4/pthread.m4 (gl_PTHREAD_CHECK):
* m4/threadlib.m4 (gl_THREADLIB_BODY): Prefer -pthread to -lpthread.
* m4/pthread_sigmask.m4 (gl_FUNC_PTHREAD_SIGMASK):
Treat -pthread like -lpthread.

10 years agoerror: drop spurious semicolon
Eric Blake [Thu, 4 Sep 2014 12:26:06 +0000 (06:26 -0600)]
error: drop spurious semicolon

* lib/error.c (__error_at_line): Fix ';;'.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agognulib-common.m4: port to GCC 4.2.1 and Sun Studio 12 C++
Paul Eggert [Tue, 2 Sep 2014 18:56:55 +0000 (11:56 -0700)]
gnulib-common.m4: port to GCC 4.2.1 and Sun Studio 12 C++

* m4/gnulib-common.m4 (AC_C_RESTRICT):
Override AC_C_RESTRICT unconditionally.
Update from autoconf, incorporating:
2014-09-02 autoconf: port 'restrict' to GCC 4.2.1
2009-01-28 Fix AC_C_RESTRICT for Sun Studio 12 C++.

10 years agomanywarnings: add GCC 4.9 warnings
Paul Eggert [Mon, 1 Sep 2014 08:51:06 +0000 (01:51 -0700)]
manywarnings: add GCC 4.9 warnings

Also, make it easier to maintain this in the future.
* build-aux/gcc-warning.spec: Add -Wabi-tag,
-Wconditionally-supported, -Wdelete-incomplete,
-Winherited-variadic-ctor, -Wvirtual-move-assign, -Wzerotrip.
Remove duplicates.  Use tabs uniformly, as that's what 'cut' wants.
* m4/manywarnings.m4 (gl_MANYWARN_ALL_GCC): Add -Wdate-time,
-Wopenmp-simd.  Use -fdiagnostics-show-option and -funit-at-a-time
only for older GCC versions that need them.  Handke
-Wnormalized=nfc specially, so that the 'comm' command used
for maintenance doesn't get confused.

10 years agovasnprintf: fix bugs in width computation
Paul Eggert [Mon, 1 Sep 2014 02:19:44 +0000 (19:19 -0700)]
vasnprintf: fix bugs in width computation

* lib/vasnprintf.c (VASNPRINTF):
Rework previous change, which introduced a bug,
to avoid the warning in a different way.
Avoid undefined behavior if the width arg is less than -INT_MAX.
Avoid unnecessary use of HAS_WIDTH local.

10 years agovasnprintf: Avoid signed/unsigned comparison warning.
Thien-Thi Nguyen [Sun, 31 Aug 2014 08:27:36 +0000 (10:27 +0200)]
vasnprintf: Avoid signed/unsigned comparison warning.

* lib/vasnprintf.c (VASNPRINTF): To calculate padding,
compare end addr of generated string w/ maximum end addr.

10 years agoparse-datetime: Avoid signed/unsigned comparison warning
Mats Erik Andersson [Sat, 30 Aug 2014 08:31:35 +0000 (10:31 +0200)]
parse-datetime: Avoid signed/unsigned comparison warning

* lib/parse-datetime.y (yylex): Compare pointer to end of buffer,
instead of calculating difference of pointers.  This removes an
annoying warning, devoid of any use.

10 years agoqsort_r: new module, for GNU-style qsort_r
Paul Eggert [Fri, 29 Aug 2014 20:00:16 +0000 (13:00 -0700)]
qsort_r: new module, for GNU-style qsort_r

This works even on FreeBSD, which has an incompatible qsort_r API.
* MODULES.html.sh: Add it.
* doc/glibc-functions/qsort_r.texi: It's now supported.
* lib/qsort.c: New file, taken from glibc with minor changes
inside "#ifndef _LIBC" and with an unnecessary "#include <alloca.h>"
removed.
* lib/qsort_r.c: New file, compiled only on FreeBSD.
* lib/stdlib.in.h (qsort_r): Declare in the usual way.
* m4/stdlib_h.m4 (gl_STDLIB_H_DEFAULTS):
* modules/qsort_r, modules/qsort_r-tests: New files.
* modules/stdlib (Makefile): Set up its defaults.
* tests/test-qsort_r.c: New file.

10 years agovla: new module
Paul Eggert [Fri, 8 Aug 2014 00:40:01 +0000 (17:40 -0700)]
vla: new module

vla: new module

GNU RCS can use this, mostly for documentation I expect.  See:
http://lists.gnu.org/archive/html/bug-gnulib/2014-08/msg00025.html
* MODULES.html.sh: Add vla.
* lib/vla.h, modules/vla: New files.

10 years agolocalename: make gl_locale_name_thread really thread-safe on Windows
Daiki Ueno [Wed, 6 Aug 2014 23:46:52 +0000 (08:46 +0900)]
localename: make gl_locale_name_thread really thread-safe on Windows

* lib/localename.c [WINDOWS_NATIVE && !IN_LIBINTL]: Include
"glthread/lock.h".
(get_lcid_lock) [WINDOWS_NATIVE]: New variable.
(get_lcid) [WINDOWS_NATIVE]: Lock while looking for an LCID.

10 years agogetpass: don't assume struct termios
Paul Eggert [Fri, 8 Aug 2014 01:13:41 +0000 (18:13 -0700)]
getpass: don't assume struct termios

Problem report and trivial fix by Jonas 'Sortie' Termansen in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-08/msg00015.html
* lib/getpass.c (getpass): Port to systems lacking struct termios.

10 years agogetdtablesize: fall back on sysconf (_SC_OPEN_MAX)
Paul Eggert [Fri, 8 Aug 2014 01:04:32 +0000 (18:04 -0700)]
getdtablesize: fall back on sysconf (_SC_OPEN_MAX)

Problem reported by Jonas 'Sortie' Termansen in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-08/msg00023.html
* lib/getdtablesize.c (getdtablesize) [_SC_OPEN_MAX]:
Implement via sysconf for platforms that lack getdtablesize.

10 years agovararrays: modernize AC_C_VARARRAYS for C11
Paul Eggert [Fri, 8 Aug 2014 00:25:28 +0000 (17:25 -0700)]
vararrays: modernize AC_C_VARARRAYS for C11

This backports a change I recently made to Autoconf.
* m4/vararrays.m4 (AC_C_VARARRAYS): Define __STDC_NO_VLA__ if
VLAs are not supported, as this is what C11 does.  The old macro
HAVE_C_VARARRAYS is still defined if they are supported, but is
now obsolescent.  Also, check for VLA bug in GCC 3.4.3.

10 years agorelocatable-prog-wrapper: port gettext to OS X 10.8 + GCC 4.8.1
Alessandro Degano [Thu, 7 Aug 2014 16:52:04 +0000 (09:52 -0700)]
relocatable-prog-wrapper: port gettext to OS X 10.8 + GCC 4.8.1

* build-aux/install-reloc (func_create_wrapper): Also wrap
strerror-override, stat, stat.

10 years agosys_select: fix FD_ZERO problem on Solaris 10
Paul Eggert [Tue, 5 Aug 2014 20:19:57 +0000 (13:19 -0700)]
sys_select: fix FD_ZERO problem on Solaris 10

* lib/sys_select.in.h: Fix Solaris 10 bug where "#include
<sys/time.h>" followed by "#include <sys/select.h>" caused FD_ZERO
to expand to an expression that invoked memset without necessarily
including <string.h>.  The problem was that the first include
defined _SYS_TIME_H, causing the second include to short-circuit.
Fix a similar problem with <sys/types.h> followed by <sys/select.h>.
Also, fix what appears to be a cut-and-paste typo, by replacing
_GL_SYS_SELECT_H_REDIRECT_FROM_SYS_TIME_H with
_GL_SYS_SELECT_H_REDIRECT_FROM_SYS_TYPES_H.

10 years agoaccept: document Solaris 10 type glitch
Paul Eggert [Tue, 5 Aug 2014 20:14:04 +0000 (13:14 -0700)]
accept: document Solaris 10 type glitch

* doc/posix-functions/accept.texi (accept): Mention that
Solaris 10 'accept' takes void * last arg, not socklen_t *.

10 years agoextern-inline: port to FreeBSD, DragonFly
Paul Eggert [Mon, 4 Aug 2014 18:30:33 +0000 (11:30 -0700)]
extern-inline: port to FreeBSD, DragonFly

* m4/extern-inline.m4 (gl_EXTERN_INLINE): Also assume the bug
is present if either __DragonFly__ or __FreeBSD__ is defined.
FreeBSD problem reported by Andrey Borzenkov in:
http://lists.gnu.org/archive/html/bug-gnulib/2014-07/msg00104.html
Also, worry about __APPLE__ only if __MACH__ is also defined,
as this is more consistent with the rest of gnulib.
(_GL_EXTERN_INLINE_STDHEADER_BUG): Rename this internal macro from
_GL_EXTERN_INLINE_APPLE_BUG, since the bug is not limited to Apple.

10 years agoautoupdate
Karl Berry [Fri, 1 Aug 2014 12:07:12 +0000 (05:07 -0700)]
autoupdate

10 years agoUse consistent style to check DEBUG macro in regex_internal.c
Siddhesh Poyarekar [Fri, 1 Aug 2014 02:04:42 +0000 (07:34 +0530)]
Use consistent style to check DEBUG macro in regex_internal.c

The DEBUG macro is checked using both #if and #ifdef in
regex_internal.c.  Make this consistent with mktime, where we use "#if
defined DEBUG && DEBUG", to be compatible with gnulib as well as glibc
style.

        regex: Make #if/#ifdef usage consistent for DEBUG
        * lib/regex_internal.c: Use "#if defined DEBUG && DEBUG" instead
        of the inconsistent usage of #if and #ifdef as that works with
        both Glibc and Gnulib's style.

10 years agoopenat-die: use _Noreturn markup
Eric Blake [Wed, 30 Jul 2014 21:01:44 +0000 (15:01 -0600)]
openat-die: use _Noreturn markup

Compiling coreutils.git on cygwin with gcc 4.8.3, I got:

lib/openat-die.c:34:1: error: function might be candidate for attribute 'noreturn' [-Werror=suggest-attribute=noreturn]
 openat_save_fail (int errnum)
 ^

* modules/openat-die (Depends-on): Add snippet/_Noreturn.
* lib/openat-die.c (openat_save_fail, openat_restore_fail): Mark
_Noreturn.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agotest-open: port to cygwin, which lacks Fortify
Eric Blake [Wed, 30 Jul 2014 23:11:49 +0000 (17:11 -0600)]
test-open: port to cygwin, which lacks Fortify

On cygwin, with gcc 4.8.3, I get this compilation warning:

In file included from ../../gltests/test-open.c:35:0:
../../gltests/test-open.h:35:1: warning: always_inline function might not be inlinable [-Wattributes]
 test_open (int (*func) (char const *, int, ...), bool print)
 ^

On that platform, there are no Fortify enhancements, and the
internal macro __always_inline is always defined to turn on the
gcc attribute, contrary to our needs.  Since relying on a __
prefix is already risky, it's better to limit the workaround of
commit d6ba92ea to just glibc.

* tests/test-open.h (ALWAYS_INLINE): New macro.
(__always_inline): Don't abuse internal symbol on non-glibc.

Signed-off-by: Eric Blake <eblake@redhat.com>
10 years agolocalename: Enforce declarations before statements.
Eli Zaretskii [Sat, 19 Jul 2014 12:22:26 +0000 (21:22 +0900)]
localename: Enforce declarations before statements.

* localename.c (gl_locale_name_thread): Declare 'lcid' before the
first statement.