Re: Stable vs. Default branch

[Ben Abbott]

On Oct 3, 2011, at 11:57 PM, John W. Eaton wrote:

> On  3-Oct-2011, Rik wrote:
> 
> | So, why don't you apply your patch to the default branch and we can
> | transplant it to stable after the 3.4.3 release.
> 
> If you are going to transplant it anyway, then you might as well just
> wait until after the release and apply it to the stable branch.
> 
> My point was that I think we've been doing too much development on the
> stable branch.  Even though most of the changes have been bug fixes,
> my guess is that they aren't all simple and safe fixes and could cause
> trouble.  I've certainly applied plenty of my own patches to stable
> during this release cycle, so I'm not pointing fingers.  But for the
> future, I'd like for us to have a clearer idea of what it is exactly
> that we want the stable branch to be.  My position in the past has
> been that it should be reserved for fixing regressions only, not for
> all bug or Matlab compatibility fixes, but I remember getting some
> opposition to that idea.
> 
> So what do others think?  Is fixing regressions only too restrictive?
> It seems to me that it would help to make sure that the stable release
> really is stable.  And we would spend less time worrying about fixing
> bugs on the stable branch, because I don't expect too many
> regressions.  Then we can spend more time on the default branch and
> maybe get new major releases out more frequently.
> 
> jwe


I'm ok with the idea of regressions only for stable. Under that policy what
happens when 3.4.x ends and we begin on 3.6.x?  Is default merged into
stable? (non-trivial I expect). If only regressions go to stable, it may be
preferred to return to labeling branches by release #s and close them
off when we move on?

Ben