On Tue, Mar 11, 2008 at 05:44:12AM +0100, Bruno Haible wrote:
> Peter O'Gorman wrote:
> > Attached.
>
> Thanks. I have applied this modified patch.
Thank you!
Peter
Peter O'Gorman wrote:
> Attached.
Thanks. I have applied this modified patch.
- In the doc: there is no need to enumerate the various workarounds which
work or don't work here; this info is not interesting for the user. Just
what is the portability problem and whether gnulib fixes it or
Peter O'Gorman wrote:
> Looks like this breaks on AIX 4.3.3 :(
>
> gl_CHECK_NEXT_HEADERS fails to find the absolute file name for
> stddef.h, it finds the empty string and substitutes
>
> #include ""
>
> into the generated stdarg.h
>
> It looks like gl_CHECK_NEXT_HEADERS fails for headers that
> ping?
pong! It's still on my radar. A few other issues had higher priority in the
last few days, though. Sorry.
Bruno
ping?
http://lists.gnu.org/archive/html/bug-gnulib/2008-02/msg00148.html
Peter
On Wed, Feb 27, 2008 at 09:13:06AM -0600, Peter O'Gorman wrote:
> Ok. Will do this later.
Attached.
Peter
diff --git a/doc/posix-functions/va_copy.texi b/doc/posix-functions/va_copy.texi
index c5a012b..a2ce0e3 100644
--- a/doc/posix-functions/va_copy.texi
+++ b/doc/posix-functions/va_copy.texi
On Wed, Feb 27, 2008 at 12:35:56PM +0100, Bruno Haible wrote:
> Peter O'Gorman wrote:
> This is too risky. This approach is very likely to break platforms other
> than AIX.
>
Sigh. Ok.
> We have an replacement that works well everywhere except your
> particular version of AIX 5.1. The particula
Peter O'Gorman wrote:
> The only system that we have that does not have #include_next with
> the native compiler is aix4.3.3 with xlc.
How do you know? Have you tested cc on HP-UX, the Compaq compiler on OSF/1,
Sun cc on Linux, and whatever?
> This patch hardcodes ///usr/include/stdarg.h if the s
On Wed, Feb 27, 2008 at 05:03:01AM +0100, Bruno Haible wrote:
> Instead, how about generating the stdarg.h replacement only for
> defined _AIX && !defined __GNUC__
> and letting it look like this:
>
> #ifndef _GL_STDARG_H
> #include "/usr/include/stdarg.h"
> #ifndef va_copy
> #define va_copy(a,
Peter O'Gorman wrote:
> It looks like gl_CHECK_NEXT_HEADERS fails for headers that do not
> declare any functions. stdarg.h only defines macros :(
Yes. I think this is unavoidable.
So, we cannot use gl_CHECK_NEXT_HEADERS for stdarg.h.
Instead, how about generating the stdarg.h replacement only f
On Tue, Feb 26, 2008 at 06:06:06PM -0600, Peter O'Gorman wrote:
> > After all, if there is already a va_copy, and it disagrees with our
> > substitute, we're probably wrong. Doing this removes the need for
> > generating or substituting HAVE_VA_COPY. va_copy is required to be a
> > macro, so this
On Tue, Feb 26, 2008 at 03:46:15PM -0800, Paul Eggert wrote:
> "Peter O'Gorman" <[EMAIL PROTECTED]> writes:
>
> > +#if ! @HAVE_VA_COPY@
> > +#define va_copy(a,b) ((a) = (b))
> > +#endif
>
> Wouldn't it be simpler to do this?
>
> #ifndef va_copy
> #define va_copy(a,b) ((a) = (b))
> #endif
>
> Af
"Peter O'Gorman" <[EMAIL PROTECTED]> writes:
> +#if ! @HAVE_VA_COPY@
> +#define va_copy(a,b) ((a) = (b))
> +#endif
Wouldn't it be simpler to do this?
#ifndef va_copy
#define va_copy(a,b) ((a) = (b))
#endif
After all, if there is already a va_copy, and it disagrees with our
substitute, we're pro
On Tue, Feb 26, 2008 at 04:54:16PM -0600, Peter O'Gorman wrote:
> diff --git a/m4/stdarg.m4 b/m4/stdarg.m4
> index e8e680a..36af1ad 100644
> --- a/m4/stdarg.m4
> +++ b/m4/stdarg.m4
> --- a/modules/stdarg
> +++ b/modules/stdarg
> --- /dev/null Tue Feb 26 22:22:00 2008
> +++ lib/stdarg.in.h Tue
Hi,
We could not build GnuTLS-2.2.2 on AIX-5.1 in 64 bit mode because
va_copy requires -qlanglvl=extc99, however, with that flag compilation
failed (and many warnings were issued).
The warnings were like:
1506-784 (W) Decimal integer constant "18446744073709551615" is out of range
for any use of
15 matches
Mail list logo