Re: CVS-19981209: Patch for "unsigned HOST_WIDE_INT" on Solaris2.6

["Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>]

 > From: Dave Brolley <brolley@cygnus.com>
 > 
 > Kaveh R. Ghazi wrote:
 > 
 > >   > Actually, feel free to make the change, since you're in there anyway.
 > >  > I'm going to be away on vacation for a few weeks, so you might as
 > >  > well do it.
 > >  > Dave
 > >
 > >         No problem.  Just to double check, I should do:
 > >
 > > "s/HOST_WIDE_INT/HOST_WIDEST_INT/" in cpp*.[ch]
 > >
 > >         Right?  Also, I can nuke including machmode.h?
 > > Just okay it, and I'll check in something to do this.
 > 
 > Well, I would check each one to make sure it makes sense. A quick
 > grep reveals that there areen't that many instances. Please test it
 > using --enable-c-cpplib and make sure that fix-header stil builds
 > and works OK.
 > 
 > Thanks,
 > Dave

	Um, using --enable-c-cpplib doesn't build right now, *before* I
add any patches... 

	Here's the output on mips-sgi-irix6.2:

 > /caip/u47/ghazi/gcc-testing/egcs-CVS19981219/gcc/xgcc
 >   -B/caip/u47/ghazi/gcc-testing/egcs-CVS19981219/gcc/
 >   -B/usr/local/mips-sgi-irix6.2/bin/ -c -I. -I.  -g -I./objc
 >   -I./../../gcc -I./../../gcc/config -I../../gcc -I./../../include init.c
 > init.c:0: warning: junk on line after #undef
 > /caip/u47/ghazi/gcc-testing/egcs-CVS19981219/gcc/xgcc
 >   -B/caip/u47/ghazi/gcc-testing/egcs-CVS19981219/gcc/
 >   -B/usr/local/mips-sgi-irix6.2/bin/ -fgnu-runtime -c -o linking.o
 >   -I. -I.  -g -I./objc -I./../../gcc -I./../../gcc/config -I../../gcc
 >   -I./../../include linking.m
 > In file included from objc/Object.h:31,
 >                  from linking.m:28:
 > objc/typedstream.h:124: parse error before `*'
 > make[2]: *** [linking.o] Error 1

	Note the 'init.c:0: warning: junk on line after #undef'.  It
does this on *every* C file in the bootstrap.  Then it dies compiling
linking.m for libobjc. 

	Here is the output on alphaev5-dec-osf4.0b:

 > /usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/xgcc
 >   -B/usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/
 >   -B/usr/local/alphaev5-dec-osf4.0b/bin/ -c -I. -I.  -g -I./objc
 >   -I./../../gcc -I./../../gcc/config -I../../gcc -I./../../include init.c
 > /usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/xgcc
 >   -B/usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/
 >   -B/usr/local/alphaev5-dec-osf4.0b/bin/ -fgnu-runtime -c -o linking.o
 >   -I. -I.  -g -I./objc -I./../../gcc -I./../../gcc/config -I../../gcc
 >   -I./../../include linking.m
 > In file included from
 >   /usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/include/stdio.h:363,
 >                  from objc/typedstream.h:32,
 >                  from objc/Object.h:31,
 >                  from linking.m:28:
 > /usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/include/sys/types.h:470:
 >   parse error before `physadr'
 > /usr/var/tmp/gcc-testing/egcs-CVS19981219/gcc/include/sys/types.h:470:
 >   warning: data definition has no type or storage class
 > make[2]: *** [linking.o] Error 1

	This one doesn't emit the #undef warning, but it still dies in
libobjc.  Of course, when I bootstrap without --enable-c-cpplib, it
all works fine.  Adding my patch for HOST_WIDEST_INT has no effect in
either case.  Ie, regular bootstraps still work, --enable-c-cpplib
still fails in the way I describe above.

		--Kaveh
--
Kaveh R. Ghazi			Engagement Manager / Project Services
ghazi@caip.rutgers.edu		Icon CMT Corp.