Great, just note Garth's comment: if you make pull requests with small self-contained commits with specific fixes and changes it's much easier and quicker for us to test, approve and merge into fenics master, as opposed to one big pull request at the end where it's hard to see the consequences of all the changes.
Martin On 14 August 2015 at 12:04, Albert Ziegenhagel < albert.ziegenha...@scai.fraunhofer.de> wrote: > I forked DOLFIN, ffc and instant on Bitbucket (right after 1.6 has been > finished) and created branches ‘albert/feature-vssupport’ from ‘master’ at > > > > https://bitbucket.org/albertzi/ > > > > But I did not push all of my local changes to the repository yet. > > > > Currently I managed to get a short enough path by removing the suffix for > the temporary path that instant uses entirely. I do not see the need to use > a suffix here since the temp-folder name generated by the OS should already > be unique. > > Additionally I changed the module name of the resulting python module that > is generated by instant/ffc to be a constant string instead of the string > that contains the suffix. This reduces the paths for the C++ Compiler > dramatically since the names of temporary files/folders that are created > due to the use of CMake are much shorter. > > > > These two changes have been sufficient to make the compiler work but I was > not able to test the changes on Linux/OSX yet. > > > > Lately I was trying to get the demo and test files to work to ensure > everything is working as expected. There I am facing some other problems > with many things that are not 100% compatible with Windows. > > > > Albert > > > > > > *Von:* Martin Sandve Alnæs [mailto:marti...@simula.no] > *Gesendet:* Freitag, 14. August 2015 11:17 > *An:* Garth N. Wells <gn...@cam.ac.uk> > *Cc:* Albert Ziegenhagel <albert.ziegenha...@scai.fraunhofer.de>; > fenics@fenicsproject.org > *Betreff:* Re: [FEniCS] Interest in supporting Windows > > > > Albert: another suggestion for shortening the paths is to use only the > > first N letters of the hash digest. F.ex. N = 20 should be well on the > safe side. > > > > As Jan said, we currently do not spend our limited time on maintenance > branches, > > although it wouldn't cost anything to add an unsupported maintenance > branch with > > contributed fixes. > > > > Martin > > > > On 30 July 2015 at 10:03, Garth N. Wells <gn...@cam.ac.uk> wrote: > > > > On Wed, 29 Jul 2015 at 19:21 Albert Ziegenhagel < > albert.ziegenha...@scai.fraunhofer.de> wrote: > > Hi everyone, > > > > I managed to build Dolfin 1.5 on Windows using Visual Studio including the > (in my opinion) most important dependencies like Trilinos and PETSc. > > > > Well done, I don't imagine it was easy. > > > > > > I had to apply only minor changes to the Dolfin C++ source code and to the > CMake build files. Additionally I had to make some changes to the python > code that calls the form compiler. > > > > > > I suggest making pull requests in small chunks. > > > > Now the last problem I seem to face is that Visual C++ has a limit for > file paths of 260 characters. Because of the (quite long) hash signatures > that are included multiple times in the path to the build directories this > limit is exceeded and the Visual C++ compiler is failing to compile the > forms. In my opinion it would be sufficient to include the hash signature > only once in the path (in the name of the cache directory) and not > additionally in the names of the source files and projects. This would > dramatically reduce the length of paths Visual C++ has to deal with and > should be sufficient to make it work. > > > > > > Good point. We could work on reducing the path length. > > > > Now my question is whether there is enough interest within the FEniCS > development team to assist me in applying the necessary changes. In return > I’m definitely willing to contribute my changes back to the project, > verifying all the tests that come with FEniCS and ensure that the Windows > compatibility stays up to date in coming releases. > > > > > > We'd be happy to accept changes. Maintenance is an issue because no > developers and very few users use WIndows. Also, we also don't have > buildbot coverage for Windows so it will take some community effort to keep > Windows builds working. It would be great if you're willing to do this. > > > > Garth > > > > > > I’m looking forward to your opinions and answers. > > > > Best regards, > > Albert > > _______________________________________________ > fenics mailing list > fenics@fenicsproject.org > http://fenicsproject.org/mailman/listinfo/fenics > > > _______________________________________________ > fenics mailing list > fenics@fenicsproject.org > http://fenicsproject.org/mailman/listinfo/fenics > > >
_______________________________________________ fenics mailing list fenics@fenicsproject.org http://fenicsproject.org/mailman/listinfo/fenics
