This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] xfail i386-ssetypes-.c
> > I guess that filling PR as tree-ssa regression won't make the goal of
> > reducing number of tree-ssa regressions any better. I will try to
> > prepare the more involved patch today.
> It's your call, of course :-) But, sometimes, patches introduce more
> regressions than they fix.
<not-100%-serious> The trick is to apply busy patch to mainline, so
these won't be tree-ssa regressions. </not-100%-serious>
It is clear that Richard don't like the conservative fix
and I fully agree that the current scheme is incredibly fragile (I was
concerned about this all the time, just didn't managed to find any
testcases) so this is good motivation to dig into that.
> Eric Botcazou