Re: gawk-3.1.3-mbstate_t.patch for hpux

[Stepan Kasal]

Hello Michael,

On Wed, Dec 10, 2003 at 04:04:57PM -0500, Michael Elizabeth Chastain wrote:
> This patch improves multibyte support detection for gawk.  It's versus
> gawk 3.1.3.

thank you very much for the patch.  I have a few comments, though I'm
not the maintainer of gawk.

The gawk maintainer, Arnold Robbins, is very busy these days.
So let's just hope that he would agree with my opinion.  ;-)

> gawk really requires mbstate_t.  Look at strncasecmpmbs for example.  So
> it's reasonable to add a HAVE_MBSTATE_T test.

I think it's better to use the AC_FUNC_MBRTOWC() macro instead.
It checks whether both mbrtowc() *and* mbstate_t are available.

Of course, it's necessary to remove the generic check for mbrtowc().

With this change, there is no need to change awk.h nor regex_internal.h .

So the patch would change only the README and configure.ac, then one
would run autoconf.
I beleive there is no nee to put the following to your patch:
>       * configure: Regenerate.
>       * configh.in: Regenerate.

(But the following would be probably still necessary:
   * regex_internal.h: Include stdarg.h/varargs.h before wchar.h.
)

> After this patch, I can go after the _XOPEN_SOURCE problem.

I guess it would be enough to document (in README.hpux) how one compiles
gawk with multibyte support in on such system.  I guess that adding the
following
        #define HAVE_MBRTOWC 1
        #include <sys/_mbstate_t.h>
to the end of custom.h (which ./configure appends to the end of config.h)
should be enough.  If I'm right, you can mention it in the README.

Michael, are you willing to rework the patch according to my
suggestions and to verify the it really works on hpux?

If yes, please post the resulting patch again to bug-gawk list.
I think Arnold will pick it here and consider it before he makes the
next release.  Additionally, I'll try to keep it in my queue of gawk
patches, just in case.

Yours,
        Stepan