Re: Check for getcwd(NULL, 0) Was: Re: Bug#636568: xerces-c: Updated patch, submitted upstream

[Svante Signell <svante.signell@telia.com>]

On Wed, 2012-10-17 at 16:15 +0200, Guillem Jover wrote:
> On Wed, 2012-10-17 at 14:15:47 +0200, Svante Signell wrote:
> > On Wed, 2012-10-17 at 11:51 +0200, Pino Toscano wrote:
> > 
> > > If you assume getcwd is available, this check is pointless; if you
> > > don't, then you need to skip the check below is getcwd has not been
> > > found.
> > 
> > New version: (has to be indented?)
> 
> I srtarted commenting inline, but just posting the proposed changes as
> a whole seems easier (beware completely untested):
> 
> AC_CHECK_FUNCS([getcwd], [
>   AC_MSG_CHECKING([whether getcwd(NULL, 0) is supported])
>   AC_RUN_IFELSE([
>     AC_LANG_PROGRAM([[
>       #include <unistd.h>
>     ]], [[
>       char *result = getcwd(NULL, 0);
> 
>       if (result == NULL)
>         return 1;
>     ]])
>   ], [
>     AC_DEFINE([HAVE_GNU_GETCWD], [1],
>               [Define to 1 if you have support for 'getcwd(NULL, 0)' GNU extension])
>     AC_MSG_RESULT([yes])
>   ], [
>     AC_MSG_RESULT([no])
>   ], [
>     AC_MSG_NOTICE([cannot run test program while cross compiling])
>   ])
> ])
I knew it could be integrated with AC_CHECK_FUNCS but chose not to do
that due to the messy number of brackets. But your solution is nice,
avoiding the ac_cv_func_getcwd test, etc. Thanks! Tested with and w/o
getcwd, getcwd(NULL, 0) and cross options (hopefully).