Re: Stable vs. Default branch

[Rik]

On 10/03/2011 07:49 AM, address@hidden wrote:
> On Oct 2, 2011, at 11:01 PM, Jordi Guti?rrez Hermoso wrote:
>
>> > Since we now have an -rc0 release, does this mean the stable branch is
>> > now frozen? I think it should be and so does Rik. Just want to make
>> > sure we're all on the same page here.
>> > 
>> > - Jordi G. H.
> My impression what all bug fixes and doc improvements were to be committed to 
> stable indefinitely.
>
> Meaning I didn't expect the stable branch to be frozen after 3.4.x.
I think the stable branch needs to be frozen *during* the actual release
but not before or after; Otherwise, the release candidates won't prove
anything.  Right now, RC0 seems to compile and work on all the major
platforms.  If we change things, even for a good bug fix, we introduce new
uncertainties and the whole release candidate needs to be re-verified on
Mac Os, Linux, Cygwin, and MinGW.
> I did expect to see the default branch merged into the stable branch once we 
> are sure there will be no more releases of 3.4.x, and have moved on to 3.6.x.
>
> Is my understanding wrong?
I agree that this is what happens.  There should first be a merge of stable
=> default to catch any fixes which have been put only on the stable
branch.  These might include some documentation and bug fixes after the
3.4.3 release.  And then there needs to be a massive merge going the other
way from default => stable.  After that unsettling event I'm sure there's
work to be done to get everything to compile and pass tests again. 

There is a choice about when that event takes place.  For example, we could
do it immediately after the 3.4.3 release and then development related to
3.6 goals, such as work to make FLTK the default toolkit, would take place
on the stable branch.  Alternatively, we keep all development on the
default branch and only do the merge right before preparing 3.6 release
candidates.

--Rik