On Tue, Dec 17, 2013 at 11:30:47AM +0000, Garth N. Wells wrote: > On 2013-12-17 11:19, Martin Sandve Alnæs wrote: > >I think we should to update ufc&ufl&instant before ffc before dolfin, > >to get the code generated with the right versions.
ok, sounds good. > >Btw, I'll update the ffc test documentation right away. Should the > >regression test data be part of the release tarball? > > > > Having the regression data might be helpful in diagnosing any > problems that a user might have with an FFC installation. Agree. Let's make it part of the tarballs. -- Anders > Garth > > >Martin > >16. des. 2013 19:00 skrev "Anders Logg" <[email protected]> følgende: > > > >>I think we need release tarballs that contain everything + tags. > >>So we need: > >> > >>1. Tarballs containing everything stored at > >> > >> fenicsproject.org/pub/software/foo [1] > >> > >>2. Tarballs containing everything stored at > >> > >> bitbucket.org/fenics-project/dolfin/downloads [2] > >> > >>3. Tags containing just the repo stored at > >> > >> bitbucket.org/fenics-project/dolfin/downloads [2] > >> > >>The fenics-release script should handle this. > >> > >>I suggest we start with DOLFIN since it's most complex - or perhaps > >>FFC with the regression test checksumming... > >> > >https://bitbucket.org/fenics-project/dolfin/issue/10/nonlinearvariationalsolver-does-not-pass > >>[3] > >>>>> > > > >>>>> > > > >> > >https://bitbucket.org/fenics-project/dolfin/issue/151/resolvecompilerpaths-bug > >>[4] > >>>>> > > > >>>>> > > I think the first issue can be closed, and a new issue > >>opened > >>>>> > > (creating solver object in constructor). I don't know > >>about the > >>>>> > > status of the second issue. Can the involved parties > >>comment? > >>>>> > > > >>>>> > > >>>>> > UFC is not in good shape because it has half-made changes > >>from > >>>>> > January and some temporary member data. I made a Pull > >>Request to > >>>>> > clean this up at > >>>>> > > >>>>> > > >> https://bitbucket.org/fenics-project/ufc/pull-request/2/ [5] > >>>>> > > >>>>> > with a Pull Request for the corresponding DOLFIN change at > >>>>> > > >>>>> > > >> https://bitbucket.org/fenics-project/dolfin/pull-request/73/ [6] > >>>>> > > >>>>> > > Johannes has suggested a release on Thursday this week > >>which I think > >>>>> > > sounds good. > >>>>> > > > >>>>> > > To make the release process as smooth as possible and to > >>enable more > >>>>> > > frequent releases in the future, I suggest we take a few > >>minutes > >>>>> > > to discuss the process. In particular: > >>>>> > > > >>>>> > > In which way can we use Bitbucket to simplify the release > >>process? > >>>>> > > > >>>>> > > Which steps need to be taken (tagging, uploading, testing > >>etc)? I > >>>>> > > think we need to (re)create a cookbook for this. Remember > >>this is > >>>>> > > the first Bitbucket release we make. > >>>>> > > > >>>>> > > Is the release script (fenics-release) functional? Can it > >>be fixed? > >>>>> > > > >>>>> > > >>>>> > Not sure about it being functional, but it will need to > >>manage the > >>>>> > generated code that is no longer under version control. > >>>>> > > >>>>> > Do we want to ship the generated code in the release > >>tarball, or > >>>>> > require that a user has the whole toolchain installed? The > >>upside of > >>>>> > shipping the generated code is that a user can run C++ demos > >>without > >>>>> > FFC (although there may be some generated code inside the > >>library). > >>>>> > The downside is that we can't just tag a changeset or a > >>branch as a > >>>>> > release. I guess for Debian/Ubuntu packages it doesn't make > >>much > >>>>> > difference since demos are part of the doc package. > >>>>> > >>>>> It seems that on bitbucket you can have both. Check > >>>>> https://bitbucket.org/fenics-project/dolfin/downloads [7] > >>>>> - Tags > >>>>> - Downloads > >>>>> > >>>>> I vote for having a release-tagged master available as machine > >>specific > >>>>> scripts for installation of a current master can be simply > >>altered for > >>>>> installing the release. > >>>>> > >>>>> Jan > >>>>> > >>>>> > > >>>>> > Garth > >>>>> > > >>>>> > _______________________________________________ > >>>>> > fenics mailing list > >>>>> > [email protected] > >>>>> > http://fenicsproject.org/mailman/listinfo/fenics [8] > >>>>> > >>>>> _______________________________________________ > >>>>> fenics mailing list > >>>>> [email protected] > >>>>> http://fenicsproject.org/mailman/listinfo/fenics [8] > >>>> > >>>> > >>>> > >>>> _______________________________________________ > >>>> fenics mailing list > >>>> [email protected] > >>>> http://fenicsproject.org/mailman/listinfo/fenics [8] > >>>> > >>>_______________________________________________ > >>>fenics mailing list > >>>[email protected] > >>>http://fenicsproject.org/mailman/listinfo/fenics [8] > >>_______________________________________________ > >>fenics mailing list > >>[email protected] > >>http://fenicsproject.org/mailman/listinfo/fenics [8] > > > > > >Links: > >------ > >[1] http://fenicsproject.org/pub/software/foo > >[2] http://bitbucket.org/fenics-project/dolfin/downloads > >[3] > >https://bitbucket.org/fenics-project/dolfin/issue/10/nonlinearvariationalsolver-does-not-pass > >[4] > >https://bitbucket.org/fenics-project/dolfin/issue/151/resolvecompilerpaths-bug > >[5] https://bitbucket.org/fenics-project/ufc/pull-request/2/ > >[6] https://bitbucket.org/fenics-project/dolfin/pull-request/73/ > >[7] https://bitbucket.org/fenics-project/dolfin/downloads > >[8] http://fenicsproject.org/mailman/listinfo/fenics > > > >_______________________________________________ > >fenics mailing list > >[email protected] > >http://fenicsproject.org/mailman/listinfo/fenics _______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
