[petsc-dev] moab nightlybuild failure

[Jed Brown]

Barry Smith <bsmith at mcs.anl.gov> writes:

>   Jed,
>
>    Fine. So
>
> 1) moab.py gitcommit will always point to moab master HEAD or older
> version of moab master
>
> 2) MOAB/PETSc developers who are synchronously changing code in PETSc
> and MOAB will manually manage appropriate branches in each. They will
> not touch gitcommit in moab.py!
>
> 3) MOAB/PETSc developers who are synchronously changing code in PETSc
> and MOAB will always merge the moab branch they've developed into moab
> master before or at the same time they merge the petsc branch into
> petsc master (updating the moab.py gitcommit at the time of updating
> the petsc master from the petsc branch).
>
>    I think this clarifies things.

This sounds fine.

> What got me confused is thinking Jed saw a bigger role for gitcommit
> then it really has; which is ONLY to synchronize petsc master with
> moab master.

I thought you wanted something more advanced, and I was saying that
although we could go part way, perfect "dancing" was not going to be
automated and attempting to do so would complicate the workflow.

Note that automated testing is tricky here because it doesn't know which
branch to check out.  For this reason, it may be appropriate in cases of
new/changed APIs to wait for the MOAB branch to be merged to 'master'
before merging the corresponding PETSc branch to 'next'.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130628/69556a36/attachment.sig>