[petsc-dev] git workflow question

[Dominic Meiser]

On 07/18/2014 10:18 AM, Jed Brown wrote:
> Dominic Meiser <dmeiser at txcorp.com> writes:
>> To make things more specific, the bug fix is for DMDA's with cusp
>> vectors. I think this will be a fair amount of work and it may take a
>> while to complete. It needs work I've done in the
>> dmeiser/fix-cusp-bjacobi branch but is otherwise unrelated. I thought it
>> might be good to do it in a separate branch because of that. I wouldn't
>> want work on this bug fix to hold up things with fix-cusp-bjacobi or to
>> make that branch more difficult to review.
> If you also need features in 'master', I would branch from 'master' and
> merge 'dmeiser/fix-cusp-bjacobi'.  But doing this is risky because now
> your branch depends on the merge having been done right.  If your
> feature does not depend on more recent features in 'master', just create
> the new branch from 'dmeiser/fix-cusp-bjacobi'.
It doesn't need new features in master.
> In branch workflow, as with software development, it's good practice to
> minimize dependencies within reason.
>
> Note that 'dmeiser/fix-cusp-bjacobi' has somewhat tangled merge history,
> so if you are going to clean it up, you should do that before starting a
> new branch.  (I don't really care in this case; keeping clean topic
> branches gets easier with practice.)
I created a mess with upstream merges because I didn't appreciate the 
consequences. I read up on this and things are much clearer now. I know 
that fix-cusp-bjacobi doesn't merge into master without conflicts at 
this point (the ill conceived upstream merges were an attempt to resolve 
the conflicts in the branch). Would the best way to clean things up be 
to recreate the branch off of master? Karl, would this mess with your 
testing/reviewing of this branch?

Cheers,
Dominic

-- 
Dominic Meiser
Tech-X Corporation
5621 Arapahoe Avenue
Boulder, CO 80303
USA
Telephone: 303-996-2036
Fax: 303-448-7756
www.txcorp.com