Re: recent changes to gnulib stdio_.h break build on powerpc-ibm-aix4.3.3.0

[Bruno Haible]

Hello,

Peter O'Gorman wrote:
> include_next.m4 incorrectly deduces that this compiler understands
> #include_next. The compiler issues a warning rather than an error when
> it sees it.
> 
> This should fix:
>  --- m4/include_next.m4~ 2007-07-18 03:21:47.089858027 +0000
> +++ m4/include_next.m4  2007-07-23 14:21:26.312399370 +0000
> @@ -11,9 +11,11 @@
>    AC_LANG_PREPROC_REQUIRE()
>    AC_CACHE_CHECK([whether the preprocessor supports include_next],
>      [gl_cv_have_include_next],
> -    [AC_PREPROC_IFELSE([#include_next <stddef.h>],
> +    [save_werror_flag=$ac_c_werror_flag; ac_c_werror_flag=yes
> +       AC_PREPROC_IFELSE([#include_next <stddef.h>],
>         [gl_cv_have_include_next=yes],
> -       [gl_cv_have_include_next=no])])
> +       [gl_cv_have_include_next=no])
> +       ac_c_werror_flag=$save_werror_flag])
>    if test $gl_cv_have_include_next = yes; then
>  
>      dnl FIXME: Remove HAVE_INCLUDE_NEXT and update everything that uses it

Thanks for the info. But relying on the presence or absence of stderr output
of the compiler seems a bit risky to me. Can you rewrite the test so that
it tests whether #include_next works for real? Something like this:

  mkdir confdir1
  mkdir confdir2
  echo ... > confdir1/conftest.h
  echo ... > confdir2/conftest.h
  save_CPPFLAGS="$CPPFLAGS"
  CPPFLAGS="$CPPFLAGS -Iconfdir1 -Iconfdir2"
  AC_PREPROC_IFELSE([#include <conftest.h>], ...)
  CPPFLAGS="$save_CPPFLAGS"

? This should be more robust.

Bruno