Bob Proulx wrote:
> That seemed to mostly work. It is still failing under ia64 though. I
> think the compiler there has no support for restrict. If I define
> restrict to be nothing to disable it completely then the compilation
> succeeds.
> ...
> So apparently the configure test for a working r
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 1/27/2007 8:10 AM:
>
> Most of us are developing on sane systems with few broken or missing
> functions. If, for example, someone uses the 'string' module and stpcpy(),
> he will not get a warning: /usr/include/string.h de
Jim Meyering wrote:
> Actually restrict-detection is working just fine.
> The problem is that "config.h" isn't included.
That seemed to mostly work. It is still failing under ia64 though. I
think the compiler there has no support for restrict. If I define
restrict to be nothing to disable it co
* Paul Eggert wrote on Mon, Jan 22, 2007 at 03:22:51AM CET:
>
> My experience is that everyone who's reported a bug against SunOS
> 4.1.x for several years, is either (1) doing it only because they're
> worried we might still want to be portable to SunOS 4.1.x, or (2)
> maintaining a computer muse
Eric Blake wrote:
> > Because when cross-compiling, you cannot run "make check" -
> > it would try to execute the test programs, which always fails.
>
> An over-generalization. 'make check' generally doesn't work when
> cross-compiling, but there are instances such as compiling 64-bit using
> 32-
Jim Meyering wrote:
> anything that delays finding portability problems
> to buildbot runs is a hint that we need better process.
I fully agree; this is one of gnulib's primary goals.
The substitute now presents an opportunity to improve this.
Most of us are developing on sane systems with few
Paul Eggert wrote:
> * m4/memmem.m4 (gl_FUNC_MEMMEM): Set HAVE_DECL_MEMMEM if necessary.
> * m4/mempcpy.m4 (gl_FUNC_MEMPCPY): Set HAVE_MEMPCPY if necessary.
> ...
Some ordering constraint is missing here. We have to guarantee that the
HAVE_DECL_MEMMEM=1 statement occurs after the
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 1/27/2007 6:53 AM:
> Hi,
>
> The directories created by --create-testdir --with-tests and
> --create-megatestdir --with-tests currently don't test a lot when doing
> cross-compiles. Because when cross-compiling, you cannot
Le samedi 27 janvier 2007 à 10:52 +0100, Yoann Vandoorselaere a écrit :
> Le vendredi 26 janvier 2007 à 00:03 -0800, Paul Eggert a écrit :
> > I'd like to add a 'string' module to gnulib, which does for
> > what the sys_time module does for , namely, sets up a
> > substitute string.h that has need
Hi,
The directories created by --create-testdir --with-tests and
--create-megatestdir --with-tests currently don't test a lot when doing
cross-compiles. Because when cross-compiling, you cannot run "make check" -
it would try to execute the test programs, which always fails.
So, to detect link err
Yoann Vandoorselaere wrote:
> Attached is a patch for the strcasestr module to use generated string.h.
Thanks.
> +#if @REPLACE_STRCASESTR
Please say so if the patch is untested. It's not dishonoring to propose an
untested patch; just a warning for us that we need to read it more carefully
and th
Paul Eggert wrote:
> Here's what I actually installed.
Additionally, I change the copyright dates to include the years of the
files that were merged into lib/string_.h. Also I prefer to use 'extern' -
for clarity, and because it makes the declarations stand out better in an
editor with syntax colo
Le vendredi 26 janvier 2007 à 00:03 -0800, Paul Eggert a écrit :
> I'd like to add a 'string' module to gnulib, which does for
> what the sys_time module does for , namely, sets up a
> substitute string.h that has needed GNU declarations. That way,
> we don't need to worry about the little includ
HP-UX's cc warned about this.
"q" was set twice, but never used.
* lib/regex_internal.c (re_string_reconstruct): Remove declaration
of set-but-not-used local, "q".
Index: lib/regex_internal.c
===
RCS file: /sources/gn
14 matches
Mail list logo
