[petsc-dev] migrating repos from petsc.cs.iit.edu to bitbucket.org

Barry Smith bsmith at mcs.anl.gov
Fri Nov 16 15:51:05 CST 2012


On Nov 16, 2012, at 3:45 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> I think we could nicely do petsc-dev, petsc-release, and one buildsystem managed as a subrepo.

   I'm willing to try this.

   Barry

> 
> We could do one petsc too, bit hg branches/bookmarks kinda suck so I don't think it would be as nice.
> 
> On Nov 16, 2012 4:32 PM, "Satish Balay" <balay at mcs.anl.gov> wrote:
> to rephrase - do we do:
> 
> 1. petsc, buildsystem, petsc-release, buildsystem-release
> 2. petsc-dev, buildsystem-dev, petsc-release, buildsystem-release
> 3. petsc-dev, buildsystem, petsc-release, buildsystem-release
> 4. something else?
> 
> satish
> 
> 
> On Fri, 16 Nov 2012, Satish Balay wrote:
> 
> > Do you want to have the repo name as 'buildsystem' or
> > 'buildsystem-dev'?
> >
> > note: reclones shouldn't be need for such renames [or url changes] -
> > but one is free to do that if they want to..
> >
> > Satish
> >
> > On Fri, 16 Nov 2012, Barry Smith wrote:
> >
> > >
> > >   I am not advocating fixing all capitalizations at this time, just BuildSystems.
> > >
> > >   Plus we should tell everyone to completely reclone anyways.
> > >
> > >    Keeping the in consist BuildSystem just to prevent a small number of temporary hiccups is not a good idea. We're writing PETSc not lapack.
> > >
> > >
> > >
> > >    Barry
> > >
> > > On Nov 16, 2012, at 2:10 PM, Satish Balay <balay at mcs.anl.gov> wrote:
> > >
> > > > On Fri, 16 Nov 2012, Barry Smith wrote:
> > > >
> > > >>
> > > >> On Nov 16, 2012, at 11:14 AM, Satish Balay <balay at mcs.anl.gov> wrote:
> > > >>
> > > >>> I'm in the process of migrating repos from petsc.cs.iit.edu to
> > > >>> bitbucket.org, and have to resolve a few organization issues.
> > > >>>
> > > >>> Restrictions.
> > > >>> - all repos should be lowercase
> > > >>> - no subdir organization possible [like http://petsc.cs.iit.edu/petsc/externalpackages/]
> > > >>>
> > > >>> 1. Since we are forced to change reponame from 'BuildSystem' to
> > > >>> 'buildsystem' - I'd like to take this opportunity to change it to
> > > >>> 'buildsystem-dev' to be consistent with petsc-dev. i.e instructions
> > > >>> will be:
> > > >>>
> > > >>> hg clone https://bitbucket.org/petsc/petsc-dev
> > > >>> hg clone https://bitbucket.org/petsc/buildsystem-dev petsc-dev/config/BuildSystem
> > > >>>
> > > >>> or
> > > >>>
> > > >>> hg clone https://bitbucket.org/petsc/petsc-release
> > > >>> hg clone https://bitbucket.org/petsc/buildsystem-release petsc-release/config/BuildSystem
> > > >>>
> > > >>> and current clones would need updates to .hg/hgrc only.
> > > >>>
> > > >>> one alternative was to change the path in source to
> > > >>> petsc-dev/config/buildsystem but it will cause some breakage and other
> > > >>> inconsistencies in source file naming scheming - so I'm against that
> > > >>
> > > >>   What breakage. I really don't like having caps in one place and small letters in another. Horrible inconsistency
> > > >
> > > > I guess one can anticipate all issues and make configure deal with
> > > > them. [with autofix - or with an error message]. The couple of
> > > > issues I was thinking off:
> > > >
> > > > - different organization between petsc-release and buildsystem confusing
> > > > users.
> > > > - configure might automatically create a new 'buildsystem' clone - when
> > > > the user might still have BuildSystem - and attempt to make changes there.
> > > >
> > > > - bisection might need manual tweaking if if needs to go back and
> > > >  forth over this changepoint [from BuildSystem to buildsystem]
> > > >
> > > > Also we won't have consistant naming of 'reop' to 'file' anyway. So I
> > > > see not much motivation to preserve the 'lowercase' part of the
> > > > attribute.
> > > >
> > > > i.e buildsystem-dev -> buildsystem
> > > >    buildsystem-release -> buildsystem
> > > >
> > > > Also with this rename - lot of sourcefiles/dirs have to be renamed
> > > > [for a consistant lowercase naming scheme] - loosing annotations in
> > > > mercurial history [perhaps mercurial should handle this automatically
> > > > - but I haven't checked it in depth].
> > > >
> > > > Satish
> > > >
> > > >
> > > >
> > > >>
> > > >>    Barry
> > > >>
> > > >>>
> > > >>> another alternative is to use 'petsc', 'buildsystem', 'petsc-release',
> > > >>> 'buildsystem-release' [but we have too much baggage referring to
> > > >>> petsc-dev, with a mailing list etc reusing this name]
> > > >>>
> > > >>> 2. I plan to reorganize external package repos [that we might have
> > > >>> patches for] with a pkg prefix.
> > > >>>
> > > >>> i.e
> > > >>> http://petsc.cs.iit.edu/petsc/externalpackages/metis-5.0.2
> > > >>> http://petsc.cs.iit.edu/petsc/externalpackages/parmetis-4.0.2
> > > >>>
> > > >>> to:
> > > >>> https://bitbucket.org/petsc/pkg-metis-5.0.2
> > > >>> https://bitbucket.org/petsc/pkg-parmetis-4.0.2
> > > >>>
> > > >>> etc..
> > > >>>
> > > >>> 3. Are AMS, ctetgen, win32fe special packages and be listed at toplevel?
> > > >>> currently ctetgen is listed under 'externalpackages'. Should it be at toplevel?
> > > >>>
> > > >>> i.e
> > > >>>
> > > >>> https://bitbucket.org/petsc/win32fe-dev
> > > >>> https://bitbucket.org/petsc/ams-dev
> > > >>> https://bitbucket.org/petsc/ctetgen-dev
> > > >>>
> > > >>> Any thoughts?
> > > >>>
> > > >>> thanks,
> > > >>> Satish
> > > >>>
> > > >>>
> > > >>>
> > > >>
> > > >>
> > > >
> > >
> > >
> >
> >
> 




More information about the petsc-dev mailing list