[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: m68k in danger of being ignored for testing propagation

On Fri, Sep 23, 2005 at 05:13:30PM +0200, Bill Allombert wrote:
> > in recent weeks from the global point of view m68k has started to hold
> > up the testing migration due to both lacking buildd power and internal
> > compiler errors.  As of the most recent britney run, the number of
> > missing builds holding up the testing migration are (sorted by numbers):
> > 460 m68k
> > 227 arm
> > 174 mips
> > 148 hppa
> > 132 s390
> > 114 sparc
> > (and all others are 100 or less). I know that there is serious work being
> > put into m68k right now, but - we still have to move on, and m68k currently
> > cannot be described as catching up.  So, unless there is a serious
> > improvement on m68k happening right now, we plan to ignore m68k for testing
> > migration beginning in one week; we already have started to ignore m68k by
> > hand if that was required to allow an RC-bugfix to go in.  This change
> > doesn't mean we are dropping consideration of m68k as a release
> > architecture for etch, but of course it is a step in that direction.

> Well, the real problem here is gcc-4.0 especially at -O3. On m68k it ICE
> instead of spewing wrong code like on i386. We are very good at spotting
> ICE but poor at spotting wrong code generation which is an at least as
> serious a problem. 

Yeah, I'm pretty bad at spotting wrong code generation when the only clue
I'm given is "somewhere in the archive, -O3 may be doing the wrong thing."
:)  If you have specific knowledge that the same code which triggers ICEs on
m68k when built with -O3 also causes wrong code generation on i386, please
share it so we can warn maintainers.  If you mean that there is some other
bug in gcc-4.0 which causes wrong code generation when using -O3 in some
other circumstances, then that's bad, but I don't see how holding up testing
transitions for it is a very effective strategy if we don't even know which
code is being built wrong.

Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon@debian.org                                   http://www.debian.org/

Attachment: signature.asc
Description: Digital signature

Reply to: