BLD: fix build issues with MSVC10 on Windows. Closes gh-4245. #4892
Conversation
|
I get: at config.py line 80 |
|
right you can't access double underscore names in python, they are mangled |
|
distutils.msvccompiler.get_build_version() might work |
|
Right, forgot about that one. Why mangle a version number? I'll update to |
|
updated. skipping Travis tests, changes not covered by any tests anyway. |
|
seems to get past the point that caused problems before, now failing due to ldexp/frexp stuff, probably related to the generation of these ufuncs not this PR |
|
For reference, ldexp/frexp pr is gh-4852. |
|
So when I got the bug last time I was using Here is the last few lines, though I'm not sure it's very helpful. |
|
@mcleung yes, @juliantaylor also just noticed that. That should be a recently introduced regression. I'll create a second branch based on a commit just before that regression for you to test now. |
|
@mcleung could you try this branch: https://github.com/rgommers/numpy/tree/msvc10-fix-test? Regressions with MSVC happen once in a while unfortunately because we don't have CI set up for it. |
|
I'm just testing based on the 1.9 branch with 2.7, 3.3 and 3.4, seems to be working fine and does the same thing I always patched my local python installations too anyway so it should be safe. @rgommers can you rebase it onto the branch point, I think this should go into 1.9 possibly the commits could also be squashed. |
Note that there are a few similar patches posted to numpygh-4101 and numpygh-4245. Those were all patches to msvc9compiler in Python distutils. Monkeypatching ``MSVCCompiler.link`` is less easy than this change to ``config._check_compiler``; effect should be the same. Also updates the error message shown when initializing MSVC fails. [ci skip]
|
sure, done. |
| # flags. See issues gh-4245 and gh-4101 for details. Also | ||
| # relevant are issues 4431 and 16296 on the Python bug tracker. | ||
| from distutils import msvc9compiler | ||
| if msvc9compiler.get_build_version() >= 10: |
There was a problem hiding this comment.
I think it would be better to put this block into a try..except
newer python versions will use newer compilers and could fix the distutils issue we are working around
when this happens this could break
There was a problem hiding this comment.
I added one at first, but then removed it because a try-except block isn't going to help in that case (the code below can't really fail, and get_build_version shouldn't either). Doesn't hurt either though.
|
It installed correctly, and then when I opened python and imported numpy I got errors. A quick reboot seemed to have fixed it. A quick test shows that all seems to be working fine. |
|
k in it goes, thanks |
BLD: fix build issues with MSVC10 on Windows. Closes gh-4245.
BLD: fix build issues with MSVC10 on Windows. Closes gh-4245.
|
You may have merged the wrong one. I'm not too sure what the rebasing did. |
|
the master branch is 1.10 and currently does not compile with msvc, please try the |
|
@mcleung thanks for testing. Julian did merge the right branch. The |
|
Thanks to you guys for fixing it so quickly. I'm just doing my part reporting bugs when I find them. I'm a regular user and want to say thanks. Upgrading to 3.4 from 2.X is causing more trouble then I anticipated. Oh well, at least it's fixed now when others adopt 3.X! |
|
@mcleung thank you for the help. And it wasn't that quick, first issue opened for this was in December... |
Note that there are a few similar patches posted to gh-4101 and gh-4245. Those were all patches to msvc9compiler in Python distutils. Monkeypatching
MSVCCompiler.linkis less easy than this change toconfig._check_compiler; effect should be the same.I don't have MSVC and TravisCI tests don't help here, so needs some manual testing. I'll ping some people who have had issues with this.