This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: GCC 3.0 Release Criteria
Richard Henderson wrote:
> On Wed, May 03, 2000 at 11:04:48AM -0700, Mark Mitchell wrote:
> > I will use i586-pc-cygwin, since we already have i686 configurations.
> > (I know this makes almost no difference, but we might as well try it
> > out.)
>
> It can turn up bugs, actually. There are several aspects of
> code selection that are different and has in the past turned
> up errors.
It's true. I have one target that is "broken" when configured for
i686 becuae the lame-o assembler doesn't grok P6 opcodes. (Note to
myself: remember to make configure.in "dumb down" the triplet for this
target since GAS isn't a viable option for it.)
> If we have several x86 targets on the list, we
> might do well to distribute them across {i586,i686,k6}.
Is it enough to pick one well-exercised IA32 target and test it like
RUNTESTFLAGS='--target_board{-mi386,-mi486,-mi586,-mi686,-mk6}'?
I can easily see "no" being the answer, but it lets you hold one variable
steady instead of trying to figure out why i386-mingw works differently
than i686-cygwin (or whatever).
RJL