There is another thing that I want to mention another thing: Leif mentioned that it would be important that after installing of numpy rebuilding of sage should happen with -b, and not with -ba like I suggested. I can understand his thoughts because actually it is a good thing when sage builds the only the .pyx files which are concerned.
But -b doesn't do this out of the box you have to trick sage, because there happenend no changes to the .pyx files at all, and so the .c files don't get recompiled. One way to this is -ba which is of course more time consuming, but after upgrading from numpy-1.3.x to a higher version this gets obsolete anyway because all newer numpy versions have the same size of "flatiter". And if you build a fresh sage this wouldn't happen anyway. An other method would be touching the numpy.pxd in cython, which actually work, but this somehow interferes with the principle that one package shouldn't touch the other, and numpy isn't dependend on cython so we would artificially create one. And then we could make the changed numpy headers dependend on the concerned modules, which would be a better solution, but then we have to add ALL headers to the dependend lists because, if we upgrade to newer numpy versions, there are different or even no header changes (what is also a problem). One thing what we can do here would be put a "patched" header into the patches path which only contains a small change like an additional comment to trick sage with the timestamps and make the modules only dependend on that header, so that we only have to add one header to the entries. But these solutions don't really make me happy because they have more the taste of a bad hack then a clean solution. Has anyone a better Idea for a clean solution? Of course I will apply any change that are demanded by reviewers, but I'm actually not happy with that change, for the reasons I mentioned. I hope one can understand my concerns. -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
