[patch] egcs-ss-1998-05-02 wouldn't build on mips-sgi-irix5.2

[patch] egcs-ss-1998-05-02 wouldn't build on mips-sgi-irix5.2

[Alexandre Oliva]

[-- Attachment #1: Type: text/plain, Size: 414 bytes --]

Hi!

The latest snapshot of egcs wouldn't build on mips-sgi-irix5.2 because 
config.h would define HAVE_INTTYPES_H to `empty' instead of `1'.  So,
when compiling cccp.c, it would choke on `#if HAVE_INTTYPES_H', on
line 121.

The attached patch fixed this problem.

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

[-- Attachment #2: defines.diff --]
[-- Type: text/x-diff, Size: 590 bytes --]

  Alexandre Oliva <oliva@dcc.unicamp.br>

	* configure.in (config.h): #define to `1' instead of nothing

Index: gcc/configure.in
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/configure.in,v
retrieving revision 1.93
diff -u -r1.93 configure.in
--- gcc/configure.in	1998/04/28 05:57:03	1.93
+++ gcc/configure.in	1998/05/05 21:15:39
@@ -3016,7 +3016,7 @@
 
 	for def in `eval echo '$'$define`; do
 		echo "#ifndef $def" >>$link
-		echo "#define $def" >>$link
+		echo "#define $def 1" >>$link
 		echo "#endif" >>$link
 	done
 done

Re: [patch] egcs-ss-1998-05-02 wouldn't build on mips-sgi-irix5.2

[Jim Wilson]

I think the bug here is that the irix5 configurations have HAVE_INTTYPES_H
in xm_defines.  <inttypes.h> is not safe to use under Irix5.  If you bootstrap
using the SGI C compiler, you will get type redeclaration errors because
of conflicts between inttypes.h and sys/types.h.

Checking further, I see this is a bug introduced by Kenner's rewrite of the
mips configuration files.  xm-iris5.h originally had in it
	#undef HAVE_INTTYPES_H
Kenner then put this in xm_defines which is wrong, because that will define
it, not undefine it.  What we really need is a xm_undefines, or failing that,
some way to mark things in xm_defines which need to be undefined.  Or we
could just undo Kenner's change, and put the #undef back in xm-iris5.h.

As for the configure.in patch, 
	* configure.in (config.h): #define to `1' instead of nothing
it is harmless and may be otherwise beneficial, so we may as well accept it.

Jim