[Ksummit-2012-discuss] [ATTEND] Process issues, device model, stable
rostedt at goodmis.org
Tue Jun 19 21:19:45 UTC 2012
On Tue, 2012-06-19 at 14:09 -0700, Olof Johansson wrote:
> We normally eventually get tired of receiving late merge requests and
> start adding them to our tree explicitly as "these came in late and if
> the merge window is painless and we have time, we will send them up
> but no promises". Maybe we should just start cutting over to that mode
> sooner, but so far things haven't been _too_ bad.
I've myself have been guilty of doing late pull requests. I'll have
patches that I'll be working on and they are pretty much finished, but
I'll start working on something else. Then I notice, crap! it's getting
close to merge time, let me see what I got pending, run tests on them
and push it out.
There's usually stuff I could have pushed out earlier, but for one
reason or another, I pushed it out when I had to.
This may be another rational for having a two step process. Have a merge
window for linux-next, that could open at -rc5 and last for two weeks.
Then only what's in linux-next is allowed to go to Linus. Of course bug
fixes will always be allowed at any time.
This will make that last minute urge happen before Linus's merge. If
something horrible is noticed, then it may be found in linux-next before
it gets to Linus. Things can be pulled out of linux-next easier than
Linus's tree as Linus never rebases.
More information about the Ksummit-2012-discuss