At Fri, 6 Nov 2009 14:23:30 +0000, Mark Brown wrote:
On Fri, Nov 06, 2009 at 02:53:12PM +0100, Takashi Iwai wrote:
Ah, I thought you merged 2.6.32-rc6 to for-2.6.33 branch.
I'd like to be able to build 2.6.32 too - the build failures arrived via mainline in the merge window and are completely unrelated to ASoC, they're in the core architecture code.
Yeah, I understand, of course. I just thought build-error comes up from 2.6.33 changes.
Well, this back-merging isn't good. If needed, move to 2.6.32-rc6,
IIRC you have other fixes which are going to generate a merge up anyway when this gets merged down into your fixes branch due to other things not having being pushed to Linus yet (unless stuff got merged since I last checked).
I do wonder if it might be worth removing the fixes branches after they've been merged and starting from the last -rc (or always merging the -rcs into the fixes branches would have the same effect). That would maintain your goal of keeping the logs clean and would also improve the integration testing that the fixes get, even if that improvement should mostly be trivial.
It's fine to keep the fix branch forwarding after the merge. Feel free to do that in your side. My only point is that the backward merge from the upstream isn't regarded as a good thing.
To be honest I find it hard to see any issue with merging either way around; the difference in the logs is minor and which you want really depends on why you're looking at them and your understanding of what the branches are doing. Part of the reason I tend to do the merges this way around is that the way the topic branches are split and kept feels like a concerted effort to keep development isolated so merging up mainline so pulling mainline fixes onto the branch seems like an event.
Well, I myself don't mind much, but Linus would do. That's why I'm asking.
then merge or rebase the pending commits (just two?) to there, and give it as for-2.6.32 branch.
I've pushed the branch up as a rebase onto 2.6.32; you still get the enormous diffstat and changelog against your fix branch so not reposting it.
Thanks!
Takashi