* Ralf Wildenhues wrote on Sun, Nov 25, 2007 at 05:16:16PM CET:
I messed up.
Hehe. I did mess up, but only in *reading* git's output.
So that merge only happened in my unpublished tree, not
upstream.
Time for some coffee. Move along, nothing to see here.
Cheers,
Ralf
Hello Jim,
* Jim Meyering wrote on Mon, Dec 10, 2007 at 02:38:41PM CET:
Support versions of autoconf prior to 2.59c.
* gnulib-tool (func_emit_initmacro_done): Define m4_foreach_w
if it is not already defined.
Hi Bruno,
I've just rebased, after pulling your latest changes.
Hello Eric, all,
* Eric Blake wrote on Tue, Dec 11, 2007 at 08:18:35PM CET:
This may not be the end of the story. For M4 head, it is now triggering a
libtool bug - libtool mistakenly canonicalizes its link line, and discards
one
of the two instances of $(local_ldadd), which means we are
[ Cc:ing bug-libtool ]
Hello Bruno, Paul, all,
* Bruno Haible wrote on Sat, Dec 29, 2007 at 05:36:01PM CET:
Paul Eggert wrote:
I do have a bit of trouble reading the code, though.
It doesn't seem to match the comment: e.g., it strips a leading lt-
even when there's no /.libs/.
When I
Hello, and a Happy New Year,
quoting memmem.m4:
AC_CACHE_CHECK([whether memmem works], [gl_cv_func_memmem_works],
[AC_RUN_IFELSE([AC_LANG_PROGRAM([#include string.h],
[return !memmem (a, 1, NULL, 0);])],
[gl_cv_func_memmem_works=yes], [gl_cv_func_memmem_works=no],
Hi Eric,
* Eric Blake wrote on Sun, Jan 06, 2008 at 02:23:33AM CET:
I'd appreciate any reviews before checking it in.
Here's a rough glance at it. FWIW, the diff is not very readable
(there was a patch to diffutils out there for --more-readable).
+/* We use the Two-Way string matching
Hello,
FYI, I updated gnulib's copies of the config files, to have AIX 6.1
support.
Bruno, I suppose config.rpath and config.libpath would need adjustment
as well. Should I send a patch?
Cheers,
Ralf
Hello,
The standards.texi document uses @/ for splitting of long lines
(in URLs), this syntax requires a newer Texinfo than 4.2.
OK to apply?
Cheers,
Ralf
* DEPENDENCIES: Require Texinfo version 4.6 or newer.
diff --git a/DEPENDENCIES b/DEPENDENCIES
index ed051be..a8f9a9d 100644
---
Hi Jim,
* Jim Meyering wrote on Mon, Jan 07, 2008 at 11:34:19PM CET:
Ralf Wildenhues [EMAIL PROTECTED] wrote:
* DEPENDENCIES: Require Texinfo version 4.6 or newer.
Considering some people (like me :) are using texinfo-4.11,
and 4.6 is over 4 years old, that is a shoe-in.
Applied
* Bruno Haible wrote on Tue, Jan 08, 2008 at 03:08:17AM CET:
Ralf Wildenhues wrote:
Bruno, I suppose config.rpath and config.libpath would need adjustment
as well. Should I send a patch?
Normally, I update these files after an official libtool release. Once
a libtool-1.5.26 is released
Hello Bruno, Sam,
* Bruno Haible wrote on Wed, Jan 16, 2008 at 12:43:12PM CET:
Sam Steingold suggests to create a new gnulib module that would contain the
files
m4/libtool.m4
build-aux/ltmain.sh
As with the gettext macros, it is expected that these files get updated in
gnulib
Bruno Haible [EMAIL PROTECTED] wrote:
If you assume that A, B, C, D all have the same
size, then it's simpler (more symmetry) to write:
[...]
If you don't want to assume this, [...]
then you're looking for offsetof, not horrible code.
Cheers,
Ralf
Hello Bruno,
A couple of minor nits:
* Bruno Haible wrote on Mon, Feb 11, 2008 at 01:15:47AM CET:
bin_PROGRAMS = git-merge-changelog
git_merge_changelog_LDADD = -L. -lgnu
Please make this
git_merge_changelog_LDADD = libgnu.a
so that automake notices and adds the dependency, so that a
Hello Bruno,
* Bruno Haible wrote on Mon, Feb 11, 2008 at 01:15:47AM CET:
2008-02-10 Bruno Haible [EMAIL PROTECTED]
New module 'git-merge-changelog'.
* modules/git-merge-changelog: New file.
* lib/git-merge-changelog.c: New file.
*Thank* *you* for this, and to Jim and
* Bruno Haible wrote on Tue, Feb 12, 2008 at 12:13:41AM CET:
This creates a file
$prefix/share/dummy/javaversion.class
Surely that 'dummy' is weird, but the file should not be needed at all,
no?
I cannot reproduce this. You must have an extraneous dependency to the
module
[ Cc:ing bug-libtool, it's upstream for argz ]
Hi Mike,
* Mike Frysinger wrote on Wed, Feb 13, 2008 at 05:38:05AM CET:
the argz.m4 header checks to see if error_t is defined, but only does so by
including the argz.h header. if you try to build on a system that does
provide error_t, but not
* Mike Frysinger wrote on Wed, Feb 13, 2008 at 07:26:16PM CET:
On Wednesday 13 February 2008, Ralf Wildenhues wrote:
* Mike Frysinger wrote on Wed, Feb 13, 2008 at 05:38:05AM CET:
the argz.m4 header checks to see if error_t is defined, but only does so
by including the argz.h header
Hello Bruno,
the commit below removed some 14000 lines from the ChangeLog file.
I hope that wasn't the reason for the speedup, ;-)
but please check whether a bug in git-merge-changelog was
responsible for this.
Cheers,
Ralf
commit 65bd10479d4cb86bab4aea1424c094f8f585000c
Author: Bruno Haible
Hi Karl,
* Karl Berry wrote on Mon, Feb 25, 2008 at 06:59:39PM CET:
A Texinfo contributor made use of two argz functions that are not in the
implementation in gnulib, argz_add and argz_count. As a result, of
course compilation failed on non-glibc systems. They seemed trivial to
implement,
Hello John,
* John E. Malmberg wrote on Mon, Mar 03, 2008 at 07:01:46AM CET:
In tests/testsuite :
VMS can not deal with having a file/directory named testname and
testname.dir in the same directory.
This is because all directories are named .dir, and to be more POSIX
friendly, VMS
Hello Bruno, Simon,
* Bruno Haible wrote on Thu, Mar 06, 2008 at 12:47:31PM CET:
Simon Josefsson wrote:
I find AC_TRY_COMPILE slightly easier to work with
than AC_COMPILE_IFELSE so perhaps we could argue that it should be
non-deprecated, unless there were a strong reason to deprecate it.
Hello Bruno, Karl,
* Bruno Haible wrote on Sun, Mar 09, 2008 at 04:30:38PM CET:
Karl Berry wrote to bug-gnulib:
[EMAIL PROTECTED]bug-gnulib at gnu dot org}.
[EMAIL PROTECTED]@code{bug-gnulib at gnu dot org}}.
Personally, I don't see a reason not to use @email here. That's why
Hi Jim,
* Jim Meyering wrote on Thu, Mar 20, 2008 at 02:58:17PM CET:
Thanks! Almost perfect.
It's lacking $(srcdir) in a few places.
Oh well, I suppose since nobody except me uses VPATH builds it I get to
fix it (eventually) ...
Cheers,
Ralf
Hi Simon,
* Simon Josefsson wrote on Tue, Mar 25, 2008 at 01:47:39PM CET:
Alternatively, I could document the work-around I use to solve the
gettext vs gnulib vs autoconf problem right now: store all gnulib
generated files in git, and before running autoreconf move the gnulib
copy of
, against gnulib, should fix this. OK to apply?
Finally, the code to set $MAN in coreutils/configure.ac leaves CR
instances in the variable:
| MAN='uname.1^M chroot.1^M hostid.1^M nice.1^M who.1^M ...'
Not sure whether that may be a problem later, probably not.
Cheers,
Ralf
2008-04-06 Ralf
* Eric Blake wrote on Sun, Apr 06, 2008 at 12:26:10AM CEST:
According to Ralf Wildenhues on 4/5/2008 4:12 PM:
|
| | checking whether // is distinct from /... ls: //: No such file or
directory
| | test: =: argument expected
| | yes
|
| The patch below, against gnulib, should fix this. OK
[ eliding autoconf-patches ]
Hello Elbert,
Please don't top-post, thank you.
* Elbert Pol wrote on Sun, Apr 06, 2008 at 01:31:55PM CEST:
Thankz for the patch.
But when i use the patch and use make i get this:
cd . ash.exe U:/coreutils-6.10/build-aux/missing --run aclocal-1.10a
-I m4
Hello,
I haven't followed the discussion at all, but:
* Bruno Haible wrote on Fri, Apr 11, 2008 at 03:14:07PM CEST:
This should ideally be solved through autoconf (let autoconf try
-std=gnu99 -D__GNUC_STDC_INLINE__ instead of -std=gnu99), but since I
don't know when the next autoconf
Hi Bruno,
* Bruno Haible wrote on Wed, Apr 23, 2008 at 10:55:49PM CEST:
! LC_ALL=C tr -d '\r' t-xstrtoimax.tmp t-xstrtoimax.xo
I hear that there really do exist systems in actual use
on which tr still does not honor backslash-escapes like \r and \n.
Which platforms, please? (I use the
* Bruno Haible wrote on Sat, Apr 26, 2008 at 02:06:17PM CEST:
Ralf Wildenhues wrote:
I hear that there really do exist systems in actual use
on which tr still does not honor backslash-escapes like \r and \n.
Which platforms, please?
Solaris 10 /usr/ucb/tr. Not likely
: Likewise.
Reported by Ralf Wildenhues [EMAIL PROTECTED].
+ # Find out how to remove carriage returns from output. Solaris /usr/ucb/tr
+ # does not understand '\r'.
+ if echo solaris | tr -d '\r' | grep solais /dev/null; then
s/solais/solaris/
Likewise in the other files.
+ cr='\015
* Ralf Wildenhues wrote on Sat, Apr 26, 2008 at 06:10:56PM CEST:
+ # Find out how to remove carriage returns from output. Solaris /usr/ucb/tr
+ # does not understand '\r'.
+ if echo solaris | tr -d '\r' | grep solais /dev/null; then
s/solais/solaris/
Likewise in the other files.
Please
Hi Simon,
* Simon Josefsson wrote on Mon, Apr 28, 2008 at 04:39:08PM CEST:
--- a/m4/sys_socket_h.m4
+++ b/m4/sys_socket_h.m4
@@ -45,6 +45,7 @@ AC_DEFUN([gl_HEADER_SYS_SOCKET],
HAVE_WINSOCK2_H=0
HAVE_WS2TCPIP_H=0
else
+ AC_REQUIRE([AC_C_INLINE])
Hello Bruno,
* Bruno Haible wrote on Sun, Apr 27, 2008 at 02:11:56PM CEST:
On mingw, the test-isfinite test failed:
test-isfinite.c:169: assertion failed
FAIL: test-isfinite.exe
And after adding the same tests as for isnan(long double) - namely, that
signalling NaNs, pseudo-denormals
* Ralf Wildenhues wrote on Wed, May 28, 2008 at 09:52:30PM CEST:
Ahem, wasn't that necessary for argz_add and argz_count already,
which were added to Libtool in 2008-02-26?
OK like this? The test would have exposed the issue on, say, Solaris.
The test will skip on systems where argz.o
[ Cc:ing bug-libtool as owner of the argz module ]
Hi David,
* David Lutterkort wrote on Wed, May 28, 2008 at 09:43:13PM CEST:
I just noticed that the argz module does not contain argz_add_sep, even
though the docs indicate it does. I'd be more than happy to work up a
patch assuming the
Hi David,
* David Lutterkort wrote on Thu, May 29, 2008 at 01:06:50AM CEST:
* argz.c (argz_add_sep, argz_create, argz_create_sep, argz_replace,
argz_delete): import almost verbatim from glibc-2.7; only changes are
additional asserts and renaming of __ functions to public interface
*
* Bob Friesenhahn wrote on Tue, Jun 03, 2008 at 03:29:57AM CEST:
On Mon, 2 Jun 2008, Eric Blake wrote:
What would the libtool folks think of a patch to make libtool start
depending on gnulib, and make gnulib the master repository for the argz
replacements? I'm not convinced that diverging
Hi Jim,
* Jim Meyering wrote on Fri, Jun 13, 2008 at 09:22:31AM CEST:
Just a heads up that I'll soon be making coreutils run its gnulib-tests/
in parallel.
Cool.
Actually, the difference in total run time isn't as great as I'd hoped it
would be due to two relatively long-running tests,
* Jim Meyering wrote on Fri, Jun 13, 2008 at 10:17:36AM CEST:
Ralf Wildenhues [EMAIL PROTECTED] wrote:
Just ordering the build of some huge objects in GCC allowed to shave 30%
build time off a parallel build:
Yes, I did the same with coreutils/tests, and it made a big difference,
since
http://lists.gnu.org/archive/html/automake-patches/2008-03/msg00018.html
* Ralf Wildenhues wrote on Sat, Mar 15, 2008 at 06:28:40PM CET:
Current Autoconf allows you to do this:
AC_CONFIG_LINKS([GNUmakefile:GNUmakefile])
the only bit that's missing is that automake shouldn't remove
* Eric Blake wrote on Wed, Jun 18, 2008 at 04:01:58AM CEST:
[adding bug-texinfo, for a documentation style question]
According to Bruno Haible on 6/17/2008 7:17 PM:
|
| s/Posix/POSIX/, no?
Does anyone have a preference?
Paul had a preference for Posix and at one time changed all the
Hi Eric,
* Eric Blake wrote on Mon, Jun 23, 2008 at 02:26:16PM CEST:
--- a/tests/test-sigaction.c
+++ b/tests/test-sigaction.c
+/* Define a mask of flags required by POSIX. Some implementations
+ provide other flags as extensions, such as SA_RESTORER, that we
+ must ignore in this
OK to apply?
Or would you prefer the slightly stricter '+\{0,1\}'?
Thanks,
Ralf
2008-07-02 Ralf Wildenhues [EMAIL PROTECTED]
Recognize 'foo_LIBRARIES += libgnu.a'.
* gnulib-tool (func_emit_lib_Makefile_am): Recognize if a
makefile snippet has already specified
Hi Bruno,
* Bruno Haible wrote on Thu, Jul 03, 2008 at 12:28:09AM CEST:
The patch is right, but you need to explain the use-case:
Recognize 'foo_LIBRARIES += libgnu.a'.
No one wants to install a libgnu.a in $(foodir). Where's the module
description
that requires this patch? (I
* Bruno Haible wrote on Thu, Jul 03, 2008 at 12:53:03PM CEST:
Ralf Wildenhues wrote:
In order to avoid building the library at all by default, I was
trying a --local-dir module that consisted of
Makefile.am:
EXTRA_LIBRARIES += libgnu.a
and that worked only without
Hi Eric,
* Eric Blake wrote on Sun, Jul 13, 2008 at 05:40:51AM CEST:
In the meantime, I coded up a test; I will probably commit it as
m4/examples/append.m4 and add it to the m4 testsuite if I can get m4 sped
up. The test shows that the current behavior is indeed quadratic:
[...]
- -Dlimit=
* Paul Eggert wrote on Fri, Jul 18, 2008 at 11:24:21PM CEST:
Stepping back for a minute: a minor change like a version-number bump
should not force the system to recompile everything in sight.
Maybe there should be a way that it also should not cause an autotools
rerun?
Sorry if I'm asking
* top/GNUmakefile: Avoid 'else COND', which older GNU make
versions do not understand.
---
Hello Jim,
OK to apply?
Thanks,
Ralf
top/GNUmakefile | 26 ++
1 files changed, 14 insertions(+), 12 deletions(-)
diff --git a/top/GNUmakefile b/top/GNUmakefile
index
Hi Eric,
* Eric Blake wrote on Mon, Aug 04, 2008 at 08:42:45PM CEST:
Ralf Wildenhues Ralf.Wildenhues at gmx.de writes:
* top/GNUmakefile: Avoid 'else COND', which older GNU make
versions do not understand.
This patch causes problems:
$ make -C build distcheck
make: Entering
Hi Paolo,
* Paolo Bonzini wrote on Wed, Aug 06, 2008 at 10:28:00AM CEST:
--- a/lib/poll.c
+++ b/lib/poll.c
#ifdef _SC_OPEN_MAX
- if (nfd sysconf (_SC_OPEN_MAX))
+ static int sc_open_max = -1;
+
+ if (nfd 0
+ || (nfd sc_open_max
+ (sc_open_max != -1
+ ||
The recent coreutils sort bug related to threading made me take a look
at gnulib for similar issues. When you try to use gnulib in threaded
code, any process-global state can potentially cause problems, whether
that be static data, file descriptor state, current directory, umask,
etc. For a lot
* Paul Eggert wrote on Fri, Dec 03, 2010 at 01:39:19AM CET:
On 12/02/10 13:47, Ralf Wildenhues wrote:
* error_at_line.c ...
IMVHO it makes sense to at least document the requirement for the caller
here.
Yes. This is also a restriction for the glibc implementation, no?
Yes.
So it's
Hello,
* Bruno Haible wrote on Sun, Dec 05, 2010 at 02:12:49AM CET:
If the particular
combination of failures matches 63, 77, or 99 (as commonly used by
automake and autoconf to mean version mismatch, skip, or hard fail),
then the configure script might misbehave.
In the generated
Hello Paul,
* Paul Eggert wrote on Thu, Dec 02, 2010 at 06:52:05PM CET:
On 11/27/10 00:59, Ralf Wildenhues wrote:
Should this patch have a similar pendant for the AC_TYPE_INT64_T macro
in Autoconf?
Offhand I would say not, since that macro tests for int64_t
individually, whereas
Trying out gold (no LTO) on GCC's cc1, a log shows that the single
biggest part (20%) of user time for such a medium sized program may be
spent in SHA1 computation for the build id. I think the SHA1 code in
libiberty would benefit from an SSE implementation, such as the one
found in git or
Hello,
* Eric Blake wrote on Tue, Dec 14, 2010 at 10:14:11PM CET:
+ configmake: support more values with older automake
+ * modules/configmake (Makefile.am): Provide fallbacks for older
+ automake.
FWIW this looks safer to me because it won't regress 2.60+ setups.
Cheers,
Ralf
Hi Simon,
* Simon Josefsson wrote on Tue, Dec 14, 2010 at 08:44:02AM CET:
GNU Nettle contains optimized SHA-1, although I'm not sure it exploits
SSE operations. Generally, the crypto code in GNU Nettle is clean (no
malloc!) and leans itself to modularity, so maybe it could replace some
Hi Bruce,
* Bruce Korb wrote on Fri, Dec 31, 2010 at 12:38:10AM CET:
On my fully patched OpenSuSE 11.3 system, libposix now installs
the files listed below. What is still needed is an autoconf
macro used to detect libposix at some minimum version number.
Currently, the only annotated tag
Hello Ben, Jim,
* Ben Pfaff wrote on Mon, Jan 03, 2011 at 08:54:19PM CET:
Jim Meyering j...@meyering.net writes:
Bruce Korb wrote:
Does that git diff-index really need to be run in a separate
shell invocation?
dirty=`sh -c 'git diff-index --name-only HEAD' 2/dev/null` || dirty=
Hi Jim,
* Jim Meyering wrote on Tue, Jan 04, 2011 at 07:14:49PM CET:
In releasing coreutils-8.9, I left its gnulib submodule pointing at a
private commit that was going to be pushed, pending an ACK... Since a
slightly different commit was pushed, coreutils' submodule SHA1 was
invalid for a
Hi Bruno,
* Bruno Haible wrote on Mon, Jan 10, 2011 at 09:30:17PM CET:
2) We might remove some workarounds from gnulib, thinking that don't
affect recent systems any more. (We sometimes do this for Cygwin,
on the premise that Cygwin users update at least once every 2 years.)
In
Hello Paul,
* Paul Eggert wrote on Tue, Jan 11, 2011 at 10:17:31PM CET:
lib_SOURCES += ftoastr.h ftoastr.c dtoastr.c ldtoastr.c
Now I naively had thought that lib_SOURCES was just for
.c files that produce .o files. Is my assumption incorrect?
Yes. _SOURCES is generally for all kinds of
A small poll: can anybody confirm whether something like this:
while touch a b; mv -f a c mv -f b c wait; do :; done
can ever fail with EBUSY from rename(2), say, on an SMP system on NFS
or in some other weird setup? Can it fail on w32? Or with some Linux
kernel versions only?
Asking
* Paul Eggert wrote on Sun, Jan 23, 2011 at 09:38:58PM CET:
On 01/23/2011 04:39 AM, Eli Zaretskii wrote:
. Why are PACKAGE and VERSION needed, and where are they used?
That's another Automake thing. PACKAGE should be emacs,
and VERSION 24.0.50. VERSION is used in some auxiliary
Hello,
* Jim Meyering wrote on Thu, Jan 27, 2011 at 05:42:09PM CET:
Bruno Haible wrote:
# Error messages should not start with a capital letter
Why should error messages not start with a capital letter?
I think this quote from the GCS is at the origin:
Yes, I think so too.
* Bruce Korb wrote on Fri, Jan 28, 2011 at 05:28:10PM CET:
On 01/27/11 16:16, Karl Berry wrote:
(Although, switching topics a bit, I wonder if
users might like to have all such errors cast in the above format so
that next-error would do something sensible with them
Makes sense for where
* Ralf Wildenhues wrote on Sun, Jan 23, 2011 at 09:39:59AM CET:
Updated patch below. Question remains whether the whole patch
is overkill or not?
Here's the answer:
http://sourceware.org/ml/binutils/2011-01/msg00366.html
I've committed the patch as shown below now.
Thanks,
Ralf
move
* Karl Berry wrote on Fri, Jan 28, 2011 at 11:18:18PM CET:
fumble: you just fumbled this
I am looking for clarity that the second form is to be avoided and
On the contrary, standards.texi is completely clear that this is the
format to use. See the Errors node.
I agree.
* Simon Josefsson wrote on Sat, Jan 29, 2011 at 10:39:34PM CET:
Btw, there is nothing about Cray UNICOS in your text but there is Cray
platform specific stuff in the gnulib code. Does anyone know about
status of Cray architectures?
FWIW, I don't know about gnulib's level of support of UNICOS,
Hello Paul,
* Paul Eggert wrote on Sun, Jan 30, 2011 at 12:51:47AM CET:
--- a/lib/mktime.c
+++ b/lib/mktime.c
@@ -45,6 +63,13 @@
# define mktime my_mktime
#endif /* DEBUG */
+/* A signed type that is at least one bit wider than int. */
+#if INT_MAX = LONG_MAX / 2
+typedef long int
Hi Simon,
* Simon Josefsson wrote on Sun, Jan 30, 2011 at 10:37:48AM CET:
Ralf Wildenhues ralf.wildenh...@gmx.de writes:
* Simon Josefsson wrote on Sat, Jan 29, 2011 at 10:39:34PM CET:
Btw, there is nothing about Cray UNICOS in your text but there is Cray
platform specific stuff
Hello Paul,
* Paul Eggert wrote on Sun, Jan 30, 2011 at 06:45:24PM CET:
--- a/lib/mktime.c
+++ b/lib/mktime.c
@@ -64,12 +64,16 @@
# define mktime my_mktime
#endif /* DEBUG */
+/* Verify a requirement at compile-time (unlike assert, which is runtime).
*/
+#define verify(name,
* Paul Eggert wrote on Sun, Jan 30, 2011 at 08:19:09PM CET:
On 01/30/2011 10:20 AM, Ralf Wildenhues wrote:
+verify (long_int_is_wide_enough, INT_MAX == INT_MAX * (long_int) 2 / 2);
This doesn't make sense to me. If long_int has the same width as int,
then the right hand side overflows thus
Hello Albert,
* Albert Chin wrote on Tue, Feb 08, 2011 at 08:45:13PM CET:
wchar.h: wchar.in.h $(CXXDEFS_H) $(ARG_NONNULL_H) $(WARN_ON_USE_H)
$(AM_V_GEN)rm -f $@-t $@ \
{ echo '/* DO NOT EDIT! GENERATED AUTOMATICALLY! */'; \
27: sed -e
Hello gnulib developers,
I've started writing an Autotools FAQ:
http://thread.gmane.org/gmane.comp.sysutils.automake.general/12601/focus=12615
Input regarding gnulib (or any of the other packages) is welcome.
Thanks,
Ralf
* Ralf Wildenhues wrote on Sat, Feb 19, 2011 at 08:10:05PM CET:
2
Hello,
* Karl Berry wrote on Wed, Feb 23, 2011 at 01:25:11AM CET:
The minor advantage of having time stamps in the text
Minor? How else can we have an accurate version number? Doing it by
hand, or doing anything related to particular vc systems, proved
completely untenable for these
[ this is http://thread.gmane.org/gmane.comp.sysutils.autoconf.bugs/7834
from http://gcc.gnu.org/ml/gcc-patches/2011-02/msg01480.html
adding bug-gnulib; followups can elide bug-autoconf ]
* Ralf Wildenhues wrote on Thu, Feb 24, 2011 at 07:24:35AM CET:
IOW, it looks like the replacement code
a bug whereby strstr would mistakenly return NULL
* lib/str-two-way.h (two_way_short_needle): Correct off-by-one error
in period calculation.
(two_way_long_needle): Likewise.
Reported by Ralf Wildenhues, with the short needle and haystack.
* tests/test-strstr.c: Add Ralf's test case to trigger
Hi Bruno,
* Ralf Wildenhues wrote on Sun, Feb 20, 2011 at 06:56:49PM CET:
Bruno, OK to merge to gnulib as below, in Christian's name?
A friendly ping!
Thanks,
Ralf
2011-02-20 Christian Rössel ... (tiny change)
Markus Geimer ... (tiny change)
Fix OpenMP flag detection
[ moving from autoconf-patches ]
* Ralf Wildenhues wrote on Fri, Mar 04, 2011 at 09:46:00PM CET:
This patch is my current known set of updates for existing Fortran
macros. [...]
I'm syncing the part below to gnulib.
Thanks,
Ralf
Update AC_OPENMP macro for Lahey compiler on GNU/Linux
Hi Simon,
* Simon Josefsson wrote on Sat, Mar 05, 2011 at 12:22:56PM CET:
The reason an older longlong.m4 ends up in the *.tar.gz archive is
proabably that autoreconf --install (gettextize) pulls in an old
longlong.m4:
test -f ./configure || autoreconf --install
...
Copying file
Hello Bruno,
* Bruno Haible wrote on Sun, Mar 06, 2011 at 10:26:18PM CET:
1. autopoint, in order to get the gettext infrastructure,
2. gnulib-tool, in order to override some gettext infrastructure .m4 files
with newer ones,
3. AUTOPOINT=true autoreconf --install
in order to
By the way, do I understand correctly that this whole thing would
become less of an issue once the pot-primary branch of git Automake
is implemented and merged?
(Sorry for the non-progress on this; I'm swamped with other stuff.)
Thanks,
Ralf
* Bruno Haible wrote on Tue, Mar 15, 2011 at 09:16:53PM CET:
Paul Eggert wrote:
http://buildlogs.pld-linux.org/index.php?dist=tharch=athlonok=1name=gccid=78974776-f08b-4b31-a977-863793c73c22action=text
This log is inconsistent: In line 872 it shows
checking for strrchr... checking for
* Reuben Thomas wrote on Wed, Mar 16, 2011 at 05:17:44PM CET:
On 16 March 2011 16:09, Jim Meyering wrote:
Reuben Thomas wrote:
1. It doesn't upload the release tarball c. itself, it emits commands
to do so. Why?
Uploading is final.
Once you do it, you may have trouble undoing it,
I suggest you guys check out the parallel-tests driver from recent
Automake. It provides makefile variables for specifying a command
to run specific tests, plus options to that. That can be used for
valgrind.
Quoting the documentation:
For tests that match an extension `.EXT' listed in
Hello Bruno,
* Bruno Haible wrote on Sat, Apr 02, 2011 at 01:23:59AM CEST:
There's also a problem when the set of files of a module changes:
Dependencies to nonexistent files cause errors, like
make[2]: *** No rule to make target `../build-aux/c++defs.h', needed by
`c++defs.h'. Stop.
* Bruno Haible wrote on Sat, Apr 02, 2011 at 02:21:24AM CEST:
- If gnulib-tool removes the generated files stdio.h, stdlib.h, and similar,
the make process will rebuild a majority of the .o files and all
tests programs. Not much of a win, compared to running make clean.
I agree that
* Bruno Haible wrote on Sat, Apr 02, 2011 at 01:37:27AM CEST:
There's also a problem when the set of macros defined in .m4 files changes.
Then 'autoconf' complains.
$ ./gnulib-tool --create-testdir --dir=/tmp/testdir wchar
$ cd /tmp/testdir
$ configure
$ make
$ gnulib-tool
* Eric Blake wrote on Sat, Apr 02, 2011 at 12:42:07AM CEST:
I've run into problems most frequently with two types of related
changes: 1. gnulib changed the header.in.h template file for a
replacement header, but because we (intentionally?) don't have a
Makefile dependency, the header.h
* Bruno Haible wrote on Sat, Apr 02, 2011 at 02:00:56AM CEST:
The autotools go to a lot of
trouble to make it possible to rerun make and rebuild the
infrastructure without resorting to the big hammer of make *clean.
This is very useful. So gnulib breaks that now?
It's not gnulib that
Hi Bruno,
* Bruno Haible wrote on Sat, Apr 02, 2011 at 07:08:14PM CEST:
OK, let me get more constructive. For which of the following do you have
test cases in the Automake test suite already, that verify that after
./configure and make and some edits in the infrastructure, make
still works as
* Bruno Haible wrote on Sat, Apr 02, 2011 at 06:42:46PM CEST:
Ralf Wildenhues wrote:
A nice and clean solution to would require a AC_CONFIG_FILES_LAZY macro
or similar (similarly to how AC_CONFIG_HEADERS works, but it should also
work if the second macro argument modified the output file
* Bruno Haible wrote on Sat, Apr 02, 2011 at 06:42:46PM CEST:
Ralf Wildenhues wrote:
Ideally, those rebuild rules
work lazily for headers which are included often (i.e., if the contents
don't change, the file isn't updated).
For files that are rebuilt outside of Makefile rules
/wchar (Makefile.am): Likewise.
* modules/wctype-h (Makefile.am): Likewise.
Reported by Karl Berry and Ralf Wildenhues.
Hi Bruno,
* Bruno Haible wrote on Sat, Apr 02, 2011 at 06:50:57PM CEST:
So, for example, you would propose to change modules/byteswap from
[...]
But in that situation, when configuration decides that the system is
now good enough, the generated byteswap.h will stay on disk. make
will work
Hello,
* Mats E Andersson wrote on Sat, Apr 09, 2011 at 02:23:27PM CEST:
the m4 helper in `m4/getaddrinfo.m4` is incorrectly implementing
a test whether `struct sockaddr` contains a member `sa_len`.
Any BSD system must include `sys/types.h` in order for the test
code to deliver a
Hi Jim,
* Jim Meyering wrote on Wed, Apr 13, 2011 at 08:48:03PM CEST:
This script segfaults on opensolaris 5.11 snv_134 i86pc:
Could you (or somebody else) report this upstream, please?
Or ping me, so I'll eventually do it?
Thanks,
Ralf
Hi Bruno,
* Bruno Haible wrote on Sat, Apr 16, 2011 at 12:10:41PM CEST:
Maybe we should advertise this more? That gnulib is not only for programmers
but also contains a test suite for part of POSIX. For example, we could
mention
this in the documentation, on the web pages, and upload on ftp
401 - 500 of 677 matches
Mail list logo