[3.4 - BIB] New loop optimizer, part 1
Geoff Keating
geoffk@geoffk.org
Wed Oct 16 11:56:00 GMT 2002
Stan Shebs <shebs@apple.com> writes:
> David Edelsohn wrote:
>
> >>>>>>Zack Weinberg writes:
> >>>>>>
> >
> >Zack> The criteria for BIB are the same as the criteria for stage 2
> >Zack> development. I personally do not think that a patch of this magnitude
> >Zack> qualifies.
> >
> > Do you notice how many failures have been introduced on the trunk
> >which is in Stage 3 development? Those only have been caught and tracked
> >down due to regular bootstraps of the trunk on various targets. The Basic
> >Improvements Branch is not receiving even that level of coverage, so it is
> >likely that regressions have been introduced.
> >
> Presumably there should be something like a mini Stage 3 for b-i-b
> before it gets pulled into the trunk. Since the trunk will be in Stage 1
> at that point, some regressions are OK, but it seems like it would be
> easier to identify and fix on b-i-b first.
I believe that part of the rules for merging in from branches are that
the to-be-applied patch (that is, the merge) should be tested on at
least three targets.
--
- Geoffrey Keating <geoffk@geoffk.org>
More information about the Gcc-patches
mailing list