Stage 3 and current breakage

[Zack Weinberg]

Andrew Pinski <apinski@apple.com> writes:

> I would request a week or two of stage 3 starting today because the
> number of regressions is just huge.  Even though this will only give
> us one week for the rest of stage 2 but we can then push that
> back. The main reason why I am asking for because bootstrap on powerpc
> -apple-darwin is broken and has been broken on and off for the last
> three weeks (mainly on the weekends when I do my non-work related GCC
> work).
>
> Note stage 3 in my mind does not mean optimization can be
> added/removed, this means all of the LNO merging should stop during
> this period.  If we do not fix some of these bugs now, we would just
> push them further back.

I'd like to endorse this general idea.  Alternating short periods of
stage 2 and 3 would allow me (and I doubt I'm alone in this boat) to
put time into fixing regressions now, rather than needing to ignore
them and concentrate on the features I was supposed to wedge into 3.5.
And as we all know, the longer bugs go unfixed, the harder it tends to
be to fix them.

zw