Re: [Qemu-devel] [PULL 6/6] MAINTAINERS: Remove obsolete stable branches

[Michael Roth]

Quoting Michael S. Tsirkin (2017-01-10 13:28:44)
> On Tue, Jan 10, 2017 at 12:17:26PM -0600, Michael Roth wrote:
> > Quoting Michael S. Tsirkin (2017-01-10 11:43:24)
> > > On Tue, Jan 10, 2017 at 11:27:58AM -0600, Michael Roth wrote:
> > > > Quoting Paolo Bonzini (2017-01-10 10:45:02)
> > > > > 
> > > > > 
> > > > > On 10/01/2017 17:34, Michael Roth wrote:
> > > > > > Quoting Thomas Huth (2017-01-10 01:54:40)
> > > > > >> On 10.01.2017 04:32, Michael S. Tsirkin wrote:
> > > > > >>> On Thu, Nov 10, 2016 at 11:11:43AM +0100, Thomas Huth wrote:
> > > > > >>>> There are only very old and orphaned stable branches listed
> > > > > >>>> in the MAINTAINERS file - so this section is pretty useless
> > > > > >>>> nowadays. Let's remove it.
> > > > > >>>>
> > > > > >>>> Reviewed-by: John Snow <address@hidden>
> > > > > >>>> Signed-off-by: Thomas Huth <address@hidden>
> > > > > >>>
> > > > > >>>
> > > > > >>> Could you add some kind of entry for stable though?
> > > > > >>> Otherwise people won't know which address to CC.
> > > > > >>
> > > > > >> I think that's a question to Michael Roth - whether such an entry 
> > > > > >> should
> > > > > >> be added and how it should look like.
> > > > > > 
> > > > > > Hmm, not sure how to make this very readable. Something like this
> > > > > > perhaps?
> > > > > > 
> > > > > > Stable branches
> > > > > > ---------------
> > > > > > Current M.(N-1) Stable Branch (e.g. v2.8 if latest M.N.0 release 
> > > > > > was v2.9.0)
> > > > > > M: Michael Roth <address@hidden>
> > > > > > L: address@hidden
> > > > > > T: git git://git.qemu-project.org/qemu.git stable-M.(N-1)
> > > > > > S: Supported
> > > > > 
> > > > > I would just say
> > > > > 
> > > > > Stable branches
> > > > > ---------------
> > > > > Current Stable Branch (e.g. v2.8 if latest M.N.0 release was v2.9.0)
> > > > > M: Michael Roth <address@hidden>
> > > > > L: address@hidden
> > > > > S: Supported
> > > > > 
> > > > > Is the stable git tree used in practice, except in the days right 
> > > > > before
> > > > > a release?
> > > > 
> > > > Generally not on my end, but it's possible downstreams make use of them
> > > > as a means to just grab the latest stable release (though I don't know 
> > > > if
> > > > this is actually the case or not). I agree that having it in MAINTAINERS
> > > > isprobably of limited usefulness though.
> > > > 
> > > > I *would* like to have a fairly consistent tree that's updated more 
> > > > regularly
> > > > that people could target for backports and such, but that would probably
> > > > be via my stable-N.M-staging tree on github. Perhaps I should point
> > > > people there instead? Unfortunately that still requires some amount of
> > > > variable substitution.
> > > 
> > > Do you need multiple trees? Why not have branches?
> > 
> > Sorry, by "tree" I meant branch, e.g.:
> > 
> >   https://github.com/mdroth/qemu/tree/stable-2.7-staging
> >   https://github.com/mdroth/qemu/tree/stable-2.6-staging
> 
> I would say
> 
> T: git git://git.qemu-project.org/qemu.git stable-M.(N-1)
> 
> conveys that clearly.

Agreed as far as the formatting, but thinking more on Paolo's query I'm not
sure the qemu.org branch is preferrable over my staging tree on github.

The main issue is that I generally only update the qemu.org branches as
part of the actual releases, whereas my staging tree is updated a bit
more regularly (and in that sense more akin to the various other
development trees referenced in MAINTAINERS).

I don't use the git.qemu.org stable branches for staging because I regularly
force-push prior to release to handle reverts/fixes to avoid diverting too
much from x.y.0 with various fix-up patches from my end, and exposing that
part of the process doesn't really sit well with me for an "official" stable
branch where I think there's some implied expectation of stability (for both
code and history). I also don't really like the idea of force-pushing anything
to git.qemu.org in general as it seems like somewhat of a liability.

Would it be appropriate to reference 2 trees?

T: git git://git.qemu-project.org/qemu.git stable-M.N (releases only)
T: git git://github.com/mdroth/qemu.git stable-M.N-staging

> 
> 
> > etc. Or were you suggesting something else?
> > 
> > I don't think I can do something like a single stable-current-staging
> > branch since there's often some overlap between stable releases for N
> > and N-1 cycles, and the N branches aren't direct descendants of N-1
> > (not sure if that's what you were suggesting).
> > 
> > > 
> > > > > 
> > > > > Paolo
> > > > > 
> > > > > >>
> > > > > >>  Thomas
> > > > > >>
> > > > > >>
> > > > > >>>> ---
> > > > > >>>>  MAINTAINERS | 22 ----------------------
> > > > > >>>>  1 file changed, 22 deletions(-)
> > > > > >>>>
> > > > > >>>> diff --git a/MAINTAINERS b/MAINTAINERS
> > > > > >>>> index d8575ab..4a60579 100644
> > > > > >>>> --- a/MAINTAINERS
> > > > > >>>> +++ b/MAINTAINERS
> > > > > >>>> @@ -1574,28 +1574,6 @@ F: tcg/tci/
> > > > > >>>>  F: tci.c
> > > > > >>>>  F: disas/tci.c
> > > > > >>>>  
> > > > > >>>> -Stable branches
> > > > > >>>> ----------------
> > > > > >>>> -Stable 1.0
> > > > > >>>> -L: address@hidden
> > > > > >>>> -T: git git://git.qemu-project.org/qemu-stable-1.0.git
> > > > > >>>> -S: Orphan
> > > > > >>>> -
> > > > > >>>> -Stable 0.15
> > > > > >>>> -L: address@hidden
> > > > > >>>> -T: git git://git.qemu-project.org/qemu-stable-0.15.git
> > > > > >>>> -S: Orphan
> > > > > >>>> -
> > > > > >>>> -Stable 0.14
> > > > > >>>> -L: address@hidden
> > > > > >>>> -T: git git://git.qemu-project.org/qemu-stable-0.14.git
> > > > > >>>> -S: Orphan
> > > > > >>>> -
> > > > > >>>> -Stable 0.10
> > > > > >>>> -L: address@hidden
> > > > > >>>> -T: git git://git.qemu-project.org/qemu-stable-0.10.git
> > > > > >>>> -S: Orphan
> > > > > >>>> -
> > > > > >>>>  Block drivers
> > > > > >>>>  -------------
> > > > > >>>>  VMDK
> > > > > >>>> -- 
> > > > > >>>> 1.8.3.1
> > > > > >>>>
> > > > > >>
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > 
> > > 
>