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

[Kaveh R. Ghazi]

 > From: Jeffrey A Law <law@hurl.cygnus.com>
 > 
 > Getting hwint.h included in the config.h/hconfig.h files is
 > trivial, we should just do that from the start.  No need for a 2
 > step phase-in of that change.

	Okay.


 >   > 	Then since inttypes.h is an autoconf check on a OS header
 >   > file, we put HOST_WIDEST_INT in system.h.  Since HOST_PTR_PRINTF is
 >   > also an autoconf based check on libc's printf, it'll move to system.h
 >   > too.  That'll clean most of the extraneous cruft out of machmode.h.
 > Actually, HOST_WIDE_INT (and friends) can also be autoconf'd, which would
 > allow us to kill more of the stuff in the xm files.  I just haven't gotten
 > around to working out the details, particularly for canadian crosses.
 > 
 > [ Eventually the xm files should disappear, except for those corresponding
 >   to hosts which can't run autoconf. ]


	Are you suggesting we would use 8*AC_CHECK_SIZEOF(foo)?

Hmm, is the stage1 compiler's notion of the sizeof int/long, etc, always
the same as gcc's?  E.g. is that driven solely by the chip HW or does a
compiler have any discretion on implementation?

Also, that test calls AC_TRY_RUN, so what would we do for cross
compilers?





 >   > 	Here's a potential hwint.h I was toying with last month.  What
 >   > do you think?
 > Seems quite reasonable to me.  
 > jeff

	Okay, I'll polish it and check hwint.h in.  A patch to have
{h}config.h include it and to have system.h define HOST_WIDEST_INT
will follow.

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