Re: Please giveback crawl on sparc
On Sun, 9 Jun 2013 18:02:31 +0200, Kurt Roeckx wrote:
> On Sun, Jun 09, 2013 at 05:58:22PM +0200, Julien Cristau wrote:
> > On Sun, Jun 9, 2013 at 00:22:50 -0700, Vincent Cheng wrote:
> >
> > > Dear wanna-build team,
> > >
> > > Please give back crawl on sparc, reason being that it hits an ICE only
> > > on sparc, not on any other arch. Thanks!
> > >
> > What makes you think that ICE will magically go away?
>
> Because stadler and sompek are known to generate ICE that don't
> seem to have anything to do with gcc itself. It just that it's
> ussually gcc that gets those random errors.
Actually, this one is a genuine gcc bug. I did bother our qemu guys
for quite a while, and with their help, managed to get an emulated
setup working (should be easier with the new version in unstable).
Even then, qemu-system-sparc is extremely slow, so it takes nearly a
week(!) to do a single test build of crawl.
And I got this ICE with gcc-4.6. I also ran a build with gcc-4.7 in
parallel on another core -- it succeded.
I'm sorry but in this particular case I'm not going to submit a bug
to gcc-4.6: it's hopefully going away, and the very concept of running
multidelta on something that slow is laughable.
Somehow crawl attracts an enormous share of ICEs, there's one on Ubuntu
with gcc-4.8 on powerpc too. Thus, crawl 0.12.2-1, uploaded yesterday,
lowers optimization on all architectures I'm not able to reasonably
debug.
So this one is sort of handled already.
--
ᛊᚨᚾᛁᛏᚣ᛫ᛁᛊ᛫ᚠᛟᚱ᛫ᚦᛖ᛫ᚹᛖᚨᚲ
Reply to: