We should deprecate and eventually remove --build. We tried this previously in #908 and ended up reverting it, however looking over our reasons for reverting it all those reasons seem better solved by the randomized build directories (which we weren't using at the time) or the TMPDIR environment variable.
dstufft
All 3 comments
In #4242 @benjaminp is using this flag to undo the move to a randomized build directory. This allows him to get reproducible build output even for packages that are generating files that bake in the build directory in some way (for example Numpy). He's using this instead of addressing the problem in each project that does that to provide a more systemic solution.
dstufft
on 2 Apr 2017
In #4242 @benjaminp is using this flag to undo the move to a randomized build directory
This should still be possible. I am interested in the --build feature (which is broken, see #804 and #4242) for ccache to speed up compilations of C extensions.
jdemeyer
on 22 Aug 2017
We are having issues with this on openSUSE. The issue is that pyc files contain the build directory, at least for wheels, which is random each time. One option would be to support the -d option of compileall, which allows users to override the directory that is stored in the pyc file. This would allow us to specify the final destination directory rather than where it was built.
toddrme2178
on 21 Jun 2018
Related issues
pradyunsg
路
3Comments
nathan130200
路
3Comments
lululukas
路
3Comments
kkom
路
3Comments
GregBorrelly
路
3Comments
Most helpful comment
We are having issues with this on openSUSE. The issue is that pyc files contain the build directory, at least for wheels, which is random each time. One option would be to support the
-doption ofcompileall, which allows users to override the directory that is stored in the pyc file. This would allow us to specify the final destination directory rather than where it was built.