On Sat, Sep 13, 2014 at 12:23 AM, Balay, Satish <[email protected]> wrote:
> Some additional thoughts. > > WRT: bitbubket RFEs would be > > 1. Config to remove merge button and suggest a copy/pasteable git fetch cmd > > 2. Config to suggest only maint and master branches for pull requests. > > And on our(integrator) side we create a better tractable branch name - > perhaps: > > Pull#-integrator/contributor/feature (or a better encoding) > I like these suggestions. Matt > Satish > ------------------------------ > From: Balay, Satish <[email protected]> > Sent: 9/13/2014 12:35 AM > To: Smith, Barry F. <[email protected]>; For users of the development > version of PETSc <[email protected]> > Subject: RE: [petsc-dev] bitbucket pull requests > > I don't consider (2) as an problem of pull request. The issues > mentioned are artifacts of our current workflow. > > The primarily difference between new feature by us vs by pull request is > the way feature branch is created and populated. > > By us: > > git checkout -b branch; edit; git commit > > Pull request: > > git fetch URL branch > > The remaining workflow is exactly same for both. > > the tracking issues are primarily issues with workflow. > > And the pull request contributors are required to know the workflow. I.e > know when to create new branch off maint vs master - this in turn > determines the pull request submitted for next vs master. > > Again this is the artifact of the workflow. > > We could decide not to burden contributions with understanding the > workflow - and transfer the burden to integrators. I.e if a request is a > new branch off master but it should go into maint - we (integrators) would > do an appropriate rebase. (we would do something equivalent if the > contribution came in as a patch file) > > But so far we have been requiring contributors to know the workflow. > > Satish > ------------------------------ > From: Barry Smith <[email protected]> > Sent: 9/12/2014 7:08 PM > To: For users of the development version of PETSc <[email protected]> > Subject: [petsc-dev] bitbucket pull requests > > > The Bitbucket pull request system isn’t that great for our needs. > > 1) the damn Merge button > 2) Users should always request merging to master or maint but the > integrators need to first merge to next and then check the tests and then > merge to master (and maint) several and there is not a good way of tracking > that. It would be nice if each pull request tracked if it had been merged > to next etc > > BTW: people have done a good job in the last couple of days of cleaning > up the pull requests but there is still some old stuff in there that needs > cleaning. > > > Barry > > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener
