This is the mail archive of the
mailing list for the GCC project.
Re: Merging mem-ssa into mainline
> Zdenek Dvorak wrote on 12/06/06 19:35:
> >I know such a separation is more work for you, but I think it will
> >make it much easier to fix the breakages than if you commit the whole
> > bunch of changes at once.
> So, I've been trying to separate the patch into independent units for a
> few days now, but I am having a really hard time doing it in a way that
> makes sense. The changes are so intertwined now that I need to rewrite
> significant chunks of the independent patch to disable the other parts.
> How adamant do you feel about this separation?
well, there are many reasons why separating the unrelated changes
is good idea in general:
1) it makes reviewing easier (this of course is not an issue in this
2) it makes it possible to track down which of the changes caused a bug,
thus making debugging easier
3) it makes it easier to revert one of the changes later (e.g. if we
later found out that we cannot efficiently use multiargument VDEFs)
4) It might make merging the changes into development branches easier.
5) It makes it easier to understand what and in what ways exactly
It's a bit surprising to me that it is that difficult to separate at
least some of the changes (e.g. DSE for aggregates and support for
multiparameter VDEFs, that were contributed by other people in separate
patches, removal of V_MUST_DEFs which IIRC was implemented before the
rest of the stuff on mem-ssa branch anyway).
Nevertheless, I of course do not have any authority to dictate you what
should or should not be done. If you believe that the effort needed to
create separate patches overweights the advantages, it is up to you to
decide (or if you are unsure, perhaps asking global maintainers for
their opinion would be a good idea).