HOST_WIDE_INT broken for x86 on mainline

[Alexandre Oliva]

On May 12, 2001, Neil Booth <neil@daikokuya.demon.co.uk> wrote:

> Andreas Jaeger wrote:-
>> This is AFAIK the result of two things:
>> - Alexandre's patch to fix cross-compiling from 32-bit to 64-bit
>> platforms
>> - Honza's patches to integrate x86-64 support into the mainline.
>> 
>> Together this leads to the problem (for x86-64 we currently
>> cross-compile on ia32).

> Hmm.  Theoretically, of course, those two points are independent of
> what should be the HOST_WIDE_INT on x86, so fixing them shouldn't
> affect it.

Not really.  When the compiler targets a 64-bit machine (possibly
among other 32-bit machines, as is the case of x86[_64]),
HOST_WIDE_INT must be at least 64-bits wide.  That's why we currently
get this on targets such as x86, mips-irix6 and bi-arch sparc.

The solution for the x86 case will be to get the i[4-7]86
configurations to not be able to generate x86_64 code, and create a
separate configuration that will, and only this one will require a
64-bit HOST_WIDE_INT.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me