tested on portage, works. Reproducible: Always
We should probably decide how to handle PyPy3 2.1 before we rush into this.
Created attachment 356366 [details, diff] re-based patch 2.1-distutils-fix_handling_of_executables_and_flags.patch kick things off. In sysconfig_pypy.py some lines (at tail of the hunk) for deletion were not present to be deleted so removed them from the hunk which is odd; warrants mb someone perusing, double checking it. Patch as is takes fine
Thank you for patch. In what state is considering PyPy3 now?
(In reply to David "okias" Heidelberger from comment #3) > Thank you for patch. > > In what state is considering PyPy3 now? We have eclass patches ready for testing. Someone needs to work on some ebuilds for testing that match the paths used in eclasses. Then someone needs to build them and we can test :). The following paths were used: - /usr/$(get_libdir)/pypy-X.Y and /usr/$(get_libdir)/pypy3-X.Y for install, - /usr/bin/pypy-X.Y and /usr/bin/pypy3-X.Y for exec symlink.
ok, now I have my quadcore, so. Where can I find eclass patches?
(In reply to David Heidelberger (okias) from comment #5) > ok, now I have my quadcore, so. Where can I find eclass patches? https://bitbucket.org/mgorny/gx86-working-tree/branch/pypy-2.1 Though they may be a little outdated.
And 2.2 is out in the meantime. ;)
Looking at the release notes, this one will be even more PITA than previously. Specifically, we will need to have special numpy for it.
a special numpy, and what of pypy's own internal numpypy? Well rhetorical really, it's got to be doc'd at pypy.org or in the docs that were mostly held bck from being built in the pypy ebuild itself. Can do this. Conditions apply. Revise all previous conditions.(In reply to Jana Saout from comment #7) > And 2.2 is out in the meantime. ;) and more time passes....
(In reply to Ian Delaney from comment #9) If you don't have something useful to add, please refrain from commenting.
Well version 2.1 work for me (not using it right now, but it's saved in ::ixit overlay) and I tried make 2.2 ebuild (based on working 2.1 ebuild), but compilation after X hours failed and now I can't spare too much cpu time on pypy. I'll try paste fail log when I get to my primary computer.
I officially offer to help
http://download.ixit.cz/pypy_build.log (bugzilla has limit on 1M)
Well, that is pretty useless... why did the task get killed? Out of memory? Because building 2.2.1 works fine for me. That is also not the reason for the holdup. pypy <= 2.0.x is integrated with the build system, so that python extensions (from other packages) can automatically be built against pypy if supported (e.g. make.conf PYTHON_TARGETS) and probably a few other things. So, fiddling everything into the ebuilds correctly needs the attention of a skilled developer and is more complicated than getting pypy itself built. I guess there's just not many people with the required time. I am a developer myself, but this stuff seems way too complicated for me, so I'm just being patient. :) (in the meantime my ebuild works and I have a version that doesn't fit into the system, but is usable, but since one of the main reasons to use the newer versions might be the numpy part, I agree that getting everything to work properly is going to be a PITA ;)) I agree that having to wait that long is a poor situation, but it's really far down on the priority list and the devs are doing this for free, so unless one can buy them the time required, trying to push things at this point is just unfair.
I know it's useless, that's why I didn't get futher. Anyway, that system has 8G memory, so it should be fine. If you have working 2.2 ebuild, please share it. Thank you
+ 12 Mar 2014; Patrick Lauer <patrick@gentoo.org> + +files/2.1-distutils-fix_handling_of_executables_and_flags.patch, + +pypy-2.2.1.ebuild: + Bump for #479832, patch by Ian Delaney. It's a tad heavy: [Timer] Total: --- 7382.0 s And it ignores pypy3, but upstream hasn't moved there yet (still 2.1beta), so no rush
It also ignores all that has been said on killing slots.
Old and crufty bugs gets closed
