Re: [PULL 0/7] ppc patch queue 2013-03-22

[Scott Wood <scottwood@freescale.com>]

On 03/31/2013 06:05:40 AM, Alexander Graf wrote:
> 
> On 31.03.2013, at 12:49, Gleb Natapov wrote:
> 
> > On Tue, Mar 26, 2013 at 11:37:42AM -0500, Scott Wood wrote:
> >> On 03/25/2013 08:33:12 PM, Gleb Natapov wrote:
> >>> On Tue, Mar 26, 2013 at 12:35:09AM +0100, Alexander Graf wrote:
> >>>>
> >>>> On 26.03.2013, at 00:16, Scott Wood wrote:
> >>>>
> >>>>> On 03/25/2013 05:59:39 PM, Alexander Graf wrote:
> >>>>>> On 25.03.2013, at 23:54, Scott Wood wrote:
> >>>>>>> On 03/25/2013 05:32:11 PM, Alexander Graf wrote:
> >>>>>>>> On 25.03.2013, at 23:21, Scott Wood wrote:
> >>>>>>>>> -next?  These are bugfixes, at least partially for
> >>> regressions from 3.8 (that I pointed out before the bugs were
> >>> merged!), that should go into master.
> >>>>>>>>>
> >>>>>>>>> Also, what about:
> >>>>>>>>> http://patchwork.ozlabs.org/patch/226227/
> >>>>>>>>>
> >>>>>>>>> You've got all four patches in kvm-ppc-3.9 as of a few
> >>> weeks ago -- will you be requesting a pull for that soon?
> >>>>>>>> Sigh. I guess I've screwed up the whole "let's make -next
> >>> an unusable tree and fix regressions in a separate one" workflow
> >>> again. Sorry for that.
> >>>>>>>> Since the patches already trickled into kvm's next branch,
> >>> all we can do now is to wait for them to come back through stable,
> >>> right? Marcelo, Gleb?
> >>>>>>>
> >>>>>>> Well, you can still submit that kvm-ppc-3.9 pull request. :-)
> >>>>>> I can, but nobody would pull it, as it'd create ugly merge
> >>> commits when 3.10 opens
> >>>>>
> >>>>> That's a lousy excuse for leaving bugs unfixed.
> >>>>
> >>>> I agree. So if it doesn't hurt to have the same commits in
> >>> kvm/next and kvm/master, I'd be more than happy to send another
> >>> pull request with the important fixes against kvm/master as well.
> >>>>
> >>> If it will result in the same commit showing twice in the Linus
> >>> tree in 3.10 we cannot do that.
> >>
> >> Why?
> >>
> > Because Linus distastes it and mat refuse to pull. There is a way  
> to avoid
> > such double commits: push fix to Linus tree and merge it back to  
> next.
> 
> Yes, that's the normal workflow. But what if we screw up (like I  
> did)? Does having a working 3.9 kernel win over double commits in the  
> tree? I'd say yes, but it might be best to just ask Linus directly.
> 
> Linus, I accidentally sent a pull request including fixes that were  
> meant for master for kvm/next which got accepted. Now we have those  
> commits in there. However, I would prefer if we could have them in  
> master, so that we have a known good 3.9 kernel for kvm on powerpc.
> 
> I could send another pull request against master, but that would mean  
> that after merging things back on the next merge window, there would  
> be a few duplicate commits in the history.
> 
> Do you think that's a big no-go, or would you be ok with duplicate  
> commits in case of an occasional screwup?

It doesn't look like there's much time left before 3.9 is released (rc6  
was released yesterday, and Linus said he expects rc7 to be the last),  
so could we come to a conclusion on this soon?  While I think it's  
ridiculous that "the same commit showing twice" would be a reason to  
let regressions go unfixed[1], at the very least please request a pull  
for the fourth bugfix patch, which should also go into 3.8 stable, and  
which did not go into the "next" branch (so no "duplicate commit" issue  
there).  If that doesn't make it into 3.9, it will likely never make it  
into 3.8 stable because there will be no more 3.8 stable releases at  
that point.

-Scott

[1] It doesn't help that the bugfix patches were posted almost two  
months ago, before the patches that introduced the bug were merged...