Summary: | dev-python/pycuda-2024.1 - numpy/arrayobject.h: No such file or directory | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Leonid Kopylov <leonchik1976> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | UNCONFIRMED --- | ||
Severity: | normal | CC: | 2sks05fxw, harrisl, jstein, parona, richard.yao |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: |
https://github.com/inducer/pycuda/issues/450 https://github.com/inducer/pycuda/pull/451 |
||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 930951 | ||
Attachments: |
build.log
Another build log |
Description
Leonid Kopylov
2024-05-25 07:20:48 UTC
Created attachment 894324 [details]
build.log
*** Bug 936401 has been marked as a duplicate of this bug. *** Created attachment 898117 [details]
Another build log
I have the same problem.
The issue is that numpy installs the headers to: /usr/lib/python3.11/site-packages/numpy/_core/include/numpy/arrayobject.h /usr/lib/python3.12/site-packages/numpy/_core/include/numpy/arrayobject.h However, pycuda is passing -I/usr/lib/python3.11/site-packages/numpy/core/include to gcc. Someone changed core to _core and pycuda was not informed. Disregard that last message. The real issue is that numpy 2 is not supported and the dependencies for this package do not disallow numpy 2. You can workaround this by doing `emerge -av numpy:0/0`. Then pycuda will build. However, updates will still want to install numpy 2. Yeah, the issue is well-understood already (hence blocking the tracker + see also'd upstream bug). (In reply to Sam James from comment #6) > Yeah, the issue is well-understood already (hence blocking the tracker + see > also'd upstream bug). Perhaps someone should update RDEPEND in dev-python/pycuda/pycuda-2024.1.ebuild to `<dev-python/numpy-2.0.0[${PYTHON_USEDEP}]`. Otherwise, `emerge --jobs --keep-going --with-bdeps=y -avDuN @world` will want to rebuild pycuda when updating older systems that have it and it will break. (In reply to Richard Yao from comment #7) > (In reply to Sam James from comment #6) > > Yeah, the issue is well-understood already (hence blocking the tracker + see > > also'd upstream bug). > > Perhaps someone should update RDEPEND in > dev-python/pycuda/pycuda-2024.1.ebuild to > `<dev-python/numpy-2.0.0[${PYTHON_USEDEP}]`. Otherwise, `emerge --jobs > --keep-going --with-bdeps=y -avDuN @world` will want to rebuild pycuda when > updating older systems that have it and it will break. Actually, it is not that it wants to rebuild it, but that it wants to update it, but it will have already updated numpy. In any case, I spent far too much time on this issue when updating an older machine this morning. :/ This bug is fixed in pycuda-2024.1.2 I cp pycuda-2024.1.ebuild to pycuda-2024.1.2.ebuild and it built to completion. I have not check functionality yet. |