Re: HPUX 11.00 with aCC vs. autoconf 2.57 [#8108]

[Albert Chin]

On Sun, May 18, 2003 at 02:32:17PM -0400, Thomas Dickey wrote:
> On Sun, May 18, 2003 at 02:07:48PM -0400, Keith Bostic wrote:
> > >> The problem is autoconf is inserting the lines:
> > >> 
> > >>  #ifdef __cplusplus
> > >>  #include <stdlib.h>
> > >>  #endif
> > >> 
> > >> in its test program for the AC_REPLACE_FUNCS macro, and,
> > >> with aCC on that results in the following error:
> > >
> > > that's a workaround for a defect in glibc (has been in
> > > autoconf for a couple of years).  Ostensibly to provide
> > > a prototype for exit(), it actually is used to allow
> > > glibc to tack a "throws" onto the exit() prototype.
> > 
> > Thank you.
> > 
> > Would it be reasonable to if/else it for gcc only?
> 
> Perhaps.  But I'm not one of the "maintainers", who have their own criteria.

How's this. I don't ifdef __GNUC__ as you mention glibc above.

-- 
albert chin (address@hidden)

-- snip snip
--- lib/autoconf/c.m4.orig      Sun Dec 22 22:10:46 2002
+++ lib/autoconf/c.m4   Tue Jun 24 10:26:11 2003
@@ -735,7 +742,7 @@
 done
 rm -f conftest*
 if test -n "$ac_declaration"; then
-  echo '#ifdef __cplusplus' >>confdefs.h
+  echo '#if defined(__cplusplus) && defined(__GLIBC__)' >>confdefs.h
   echo $ac_declaration      >>confdefs.h
   echo '#endif'             >>confdefs.h
 fi