On Solaris 8, I'm seeing these warnings:
"./vasnprintf.c", line 2385: warning: argument #3 is incompatible with
prototype:
"./vasnprintf.c", line 2430: warning: argument #3 is incompatible with
prototype:
"./vasnprintf.c", line 2479: warning: argument #3 is incompatible with
prototype:
"./vasnp
The test-striconveh.c test was extended on 2009-09-05 to test also the
conversion from ASCII to ISO-8859-1 with invalid input. This test fails
on Solaris 8, 9, 10 with the native (non-GNU) Solaris iconv because it
is not able to report invalid input at least in this case.
This fixes it by undefini
I just noticed that with the recent autoupdate, config.sub now has a
copyright of 2010 but a timestamp of 2009-12-30 two lines later. It
looks rather awkward. Is there any legal issue with listing a
copyright year too soon?
It seems suboptimal, but I can't imagine that it's a sub
Eric Blake byu.net> writes:
> > Reducing two invocations of AC_CHECK_HEADERS_ONCE to a single one will
> > IMO not bring large benefits.
>
> Then how about a patch in the converse direction, that ensures
> AC_CHECK_HEADERS_ONCE is called prior to any use of ac_cv_header_xxx_h,
rather
> than t
Bruno Haible clisp.org> writes:
> > gl_CHECK_NEXT_HEADERS invokes AC_CHECK_HEADERS under the hood.
>
> Well, it does, but it's not documented. It's an abstraction violation to
> exploit this knowledge (even though we do in some places).
>
> > That means several other .m4 file can probably be si
I've learned my lesson from this morning - I won't push this until it's had a
bit longer for reviews. At any rate, this introduces warn-on-use, then
demonstrates three distinct ways to use it. I am still working on the global
change of all remaining uses of GL_LINK_WARNING over to _GL_WARN_ON_
Eric Blake wrote:
> gl_CHECK_NEXT_HEADERS invokes AC_CHECK_HEADERS under the hood.
Well, it does, but it's not documented. It's an abstraction violation to
exploit this knowledge (even though we do in some places).
> That means several other .m4 file can probably be simplified:
Reducing two invo
Eric Blake wrote:
> But your fix is also incomplete. For example:
>
> > --- 306,315
> >__THROW _GL_ARG_NONNULL ((1, 2));
> > # endif
> > #elif defined GNULIB_POSIXCHECK
> > ! # undef posix_spawnattr_getflags
> > ! # define posix_spawnattr_getflags(a, b) \
Eric Blake byu.net> writes:
> I noticed these while working towards improving link-warning over to
> compile warnings.
And one more:
From: Eric Blake
Date: Thu, 31 Dec 2009 13:43:28 -0700
Subject: [PATCH 1/2] test-dup2: avoid compiler warning
A warning cropped up from the 2009-12-28 change,
I've added this to my queue.
From: Eric Blake
Date: Thu, 31 Dec 2009 08:48:16 -0700
Subject: [PATCH] doc: correct availability of cygwin 1.5.x getopt
* doc/posix-functions/optarg.texi (optarg): Cygwin supplies getopt
variables.
* doc/posix-functions/opterr.texi (opterr): Likewise.
* doc/posix-f
Eric Blake byu.net> writes:
> Huh. gl_CHECK_NEXT_HEADERS invokes AC_CHECK_HEADERS under the hood. That
> means several other .m4 file can probably be simplified:
>
Here's what I came up with:
From: Eric Blake
Date: Thu, 31 Dec 2009 14:35:32 -0700
Subject: [PATCH 1/2] spawn: yet more fixes
Eric Blake wrote:
> + * modules/arpa_inet (Makefile.am): Always build replacement
> + header.
> + * modules/ctype (Makefile.am): Likewise.
> + * modules/dirent (Makefile.am): Likewise.
> + * modules/inttypes (Makefile.am): Likewise.
> + * modules/langinfo (Makefi
Bruno Haible clisp.org> writes:
> Some further updates of the gl_REPLACE_*_H macros and comments. I think
> these gl_REPLACE_*_H macros should be kept; we don't know how the idioms
> will evolve in the future.
Nice. In writing my patch, I had noticed that some shell variables (like
WCHAR_H) we
Bruno Haible clisp.org> writes:
>
> Eric Blake wrote:
> > + * m4/sys_select_h.m4 (gl_HEADER_SYS_SELECT): Likewise.
>
> Your patch reordered the statements in such a way that
ac_cv_header_sys_select_h
> can be tested before it is computed. This fixes it.
Thanks for catching that.
Huh.
Eric Blake wrote:
> + * m4/arpa_inet_h.m4 (gl_HEADER_ARPA_INET)
> + (gl_ARPA_INET_H_DEFAULTS): Drop unneeded variable.
> + * m4/ctype.m4 (gl_CTYPE_H_DEFAULTS): Likewise.
> + * m4/isblank.m4 (gl_FUNC_ISBLANK): Likewise.
> + * m4/dirent_h.m4 (gl_REPLACE_DIRENT_H, gl_DIRE
Bruno Haible clisp.org> writes:
> Looks like some part of this patch was misapplied:
posix_spawnattr_getsigdefault
> gets a link warning twice, and some others under the wrong #elif. I'm applying
> this fix:
But your fix is also incomplete. For example:
> --- 306,315
>__THROW _GL
Eric Blake wrote:
> + * m4/sys_select_h.m4 (gl_HEADER_SYS_SELECT): Likewise.
Your patch reordered the statements in such a way that ac_cv_header_sys_select_h
can be tested before it is computed. This fixes it.
2009-12-31 Bruno Haible
* m4/sys_select_h.m4 (gl_HEADER_SYS_SELECT):
Eric Blake wrote:
> + * m4/sys_utsname_h.m4 (gl_SYS_UTSNAME_H)
> + (gl_SYS_UTSNAME_H_DEFAULTS): Likewise.
The patch you committed lacked the modification corresponding to the second
ChangeLog line. I'm adding it:
2009-12-31 Bruno Haible
* m4/sys_utsname_h.m4 (gl_SYS_UTSNA
Bruno Haible clisp.org> writes:
> Eric Blake wrote:
> > + * tests/test-signal.c (main): Enhance test.
>
> One part of this does not look right. OK to remove this?
Yes - I spotted that earlier this morning, and it was already in my queue (one
too many copy-n-paste). But if you push befor
Eric Blake wrote:
> + * tests/test-signal.c (main): Enhance test.
One part of this does not look right. OK to remove this?
*** tests/test-signal.c.origThu Dec 31 22:18:46 2009
--- tests/test-signal.c Thu Dec 31 22:18:31 2009
***
*** 110,118
#ifdef SIGXFSZ
case
Eric Blake wrote:
> * lib/spawn.in.h (posix_spawn, posix_spawnp, posix_spawnattr_init)
> (posix_spawnattr_destroy, posix_spawnattr_getsigdefault)
> (posix_spawnattr_setsigdefault, posix_spawnattr_getsigmask)
> (posix_spawnattr_setsigmask, posix_spawnattr_getflags)
> (posix_spawnattr_setflags, posix
Eric Blake wrote:
> 2009-12-31 Eric Blake
>
> + sys_times, sys_utsname: use include_next
> + * m4/sys_times_h.m4 (gl_SYS_TIMES_H): Support wrapping an existing
> + header.
> + (gl_SYS_TIMES_H_DEFAULTS): Add another variable.
> + * m4/sys_utsname_h.m4 (gl_SYS_UTSNA
Eric Blake byu.net> writes:
> I just spotted a larger logic problem - on Linux kernels between 2.6.19 and
> 2.6.22 (when utimensat existed, but rejected AT_SYMLINK_NOFOLLOW)(, we are
now
> calling lstat twice when only once is necessary. Therefore, I think the best
> course of action is to r
Jim Meyering meyering.net> writes:
> > git config user.signingkey
>
> I'd rather not, since there's no guarantee that
> it won't have changed between the tagging operation
> and whenever someone runs the announcement-generating rule.
On the other hand, we should probably kill the (unused) VC-ta
Jim Meyering meyering.net> writes:
> utimens.c: In function 'lutimens':
> utimens.c:425: error: declaration of 'st' shadows a previous local [-
Wshadow]
> utimens.c:404: error: shadowed declaration is here [-Wshadow]
>
> This patch is nearly minimal, but perhaps not ideal.
Thanks for the
Hi Eric,
Without this patch, coreutils no longer builds
when configured with --enable-gcc-warnings:
utimens.c: In function 'lutimens':
utimens.c:425: error: declaration of 'st' shadows a previous local [-Wshadow]
utimens.c:404: error: shadowed declaration is here [-Wshadow]
This patch is n
Eric Blake wrote:
> the proposed patch is okay with me, as long as you verify that none of
> your changes to gl_PREREQ_* blocks cause problems.
Verified by inspection.
Jim Meyering wrote:
> Fine by me. Thanks!
Committed and pushed.
Bruno
Bruno Haible wrote:
> I'm applying the first patch below and proposing the additional 3 patches at
> the end. Jim, Eric, ok to apply?
...
> 2009-12-31 Bruno Haible
>
> Use AC_C_INLINE where necessary.
> * m4/chdir-long.m4 (gl_PREREQ_CHDIR_LONG): Require AC_C_INLINE.
>
> Use AC_
Eric Blake wrote:
> Jim Meyering meyering.net> writes:
>
>> I've long wanted to avoid hard-coding my GPG key ID
>> in each project's cfg.mk file. The stumbling block was
>> how to derive the key ID from the tag signature.
>> I wanted to avoid relying on the content of gpgv's (gpg --verify's)
>> d
Hi Eric,
> if gl_PREREQ_whatever is changed to have no output (as you just did
> here, by changing : to AC_REQUIRE), then you created a syntax error.
Oops, you're right. I propose to use the patch below instead.
> In general, I've been starting to steer away from trivial gl_PREREQ blocks
> (th
I just noticed that with the recent autoupdate, config.sub now has a copyright
of 2010 but a timestamp of 2009-12-30 two lines later. It looks rather
awkward. Is there any legal issue with listing a copyright year too soon?
--
Eric Blake
Bruno Haible clisp.org> writes:
> AC_DEFUN([gl_PREREQ_CHDIR_LONG],
> [
> - :
> + AC_REQUIRE([AC_C_INLINE])
> ])
In general, this is dangerous. The idiom has often been:
if condition; then
gl_PREREQ_whatever
fi
which means if gl_PREREQ_whatever is changed to have no output (as you just
Jim Meyering meyering.net> writes:
> I've long wanted to avoid hard-coding my GPG key ID
> in each project's cfg.mk file. The stumbling block was
> how to derive the key ID from the tag signature.
> I wanted to avoid relying on the content of gpgv's (gpg --verify's)
> diagnostic, but it seems ot
I've long wanted to avoid hard-coding my GPG key ID
in each project's cfg.mk file. The stumbling block was
how to derive the key ID from the tag signature.
I wanted to avoid relying on the content of gpgv's (gpg --verify's)
diagnostic, but it seems other tools (at least one perl module)
do precise
The following files use 'inline' but don't require AC_C_INLINE.
lib/chdir-long.c
lib/fatal-signal.c
lib/fts.c
lib/mbchar.h
lib/mbfile.h
lib/mbiter.h
lib/mbuiter.h
lib/regcomp.c
lib/regexec.c
lib/regex_internal.c
lib/regex_internal.h
lib/stat.c
lib/u64.h
lib/wait-process
Bruno Haible wrote:
> Eric Blake wrote:
>> But yes, it is sufficient to use the
>> lighter-weight AC_C_INLINE, rather than the full-blown 'inline' module.
>> Thanks for catching that.
>
> This overkill also occurs in a number of other places. I'm applying the first
> patch below.
>
> Jim, is the se
FYI,
>From acb9c9fcc70a0f65855e9939eb1b632b3a510d8f Mon Sep 17 00:00:00 2001
From: Jim Meyering
Date: Thu, 31 Dec 2009 15:59:29 +0100
Subject: [PATCH] maint.mk: create announcement template in ~/, not in /tmp
* top/maint.mk (emit_upload_commands): Adjust.
(release-prep): Emit into ~/announce-...
Bruno Haible clisp.org> writes:
> Jim, is the second patch (at the end) below also OK to apply?
>
> 2009-12-31 Bruno Haible clisp.org>
>
> Use AC_C_INLINE instead of module 'inline' where possible.
> * m4/openat.m4 (gl_PREREQ_OPENAT): Require AC_C_INLINE.
> * modules/openat
Eric Blake wrote:
> But yes, it is sufficient to use the
> lighter-weight AC_C_INLINE, rather than the full-blown 'inline' module.
> Thanks for catching that.
This overkill also occurs in a number of other places. I'm applying the first
patch below.
Jim, is the second patch (at the end) below als
Reviewing the uses of 'inline', 'HAVE_INLINE', AC_C_INLINE against each other,
there are a couple of mismatches.
Part 1: Uses of AC_C_INLINE that are not necessary. I'm applying this:
2009-12-31 Bruno Haible
Remove unnecessary AC_C_INLINE invocation.
* m4/popen.m4 (gl_PREREQ_
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 12/31/2009 1:23 AM:
> Eric Blake wrote:
>> 2009-12-30 Eric Blake
>>
>> + fdutimensat: remove bogus dependency
>> + * modules/fdutimensat (Depends-on): Drop inline.
>>
>
> There is still a use of 'inline' in
Bruno Haible writes:
> Hi Simon,
>
> Simon Josefsson wrote:
>> The test-localename test crashes for me:
>> #0 strcmp () at ../sysdeps/i386/i686/strcmp.S:39
>> #1 0x08049840 in test_locale_name_thread () at test-localename.c:421
>> Debugging it shows that the variables passed to strcmp are NULL.
Bruno Haible writes:
> It would be nice to be able to support uselocale and newlocale on all
> platforms, but it is unrealistic: The support code would be huge, because it
> would have to override every function that is locale dependent in some sense.
>
> Therefore, on platforms where the system
Eric Blake wrote:
> 2009-12-30 Eric Blake
>
> + fdutimensat: remove bogus dependency
> + * modules/fdutimensat (Depends-on): Drop inline.
>
There is still a use of 'inline' in lib/utimens.h. Isn't this needed?
2009-12-31 Bruno Haible
* modules/fdutimensat (configure
44 matches
Mail list logo
