Version: 0.16.1-6.fc24 Tests and real programs failing there: ====================================================================== ERROR: Failure: ImportError (/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/linalg/_flapack.so: undefined symbol: sgegv_) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/lib/python2.7/site-packages/nose/loader.py", line 418, in loadTestsFromName addr.filename, addr.module) File "/usr/lib/python2.7/site-packages/nose/importer.py", line 47, in importFromPath return self.importFromDir(dir_path, fqname) File "/usr/lib/python2.7/site-packages/nose/importer.py", line 94, in importFromDir mod = load_module(part_fqname, fh, filename, desc) File "/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/spatial/__init__.py", line 94, in <module> from ._procrustes import procrustes File "/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/spatial/_procrustes.py", line 11, in <module> from scipy.linalg import orthogonal_procrustes File "/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/linalg/__init__.py", line 172, in <module> from .misc import * File "/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/linalg/misc.py", line 6, in <module> from .lapack import get_lapack_funcs File "/home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/linalg/lapack.py", line 356, in <module> from scipy.linalg import _flapack ImportError: /home/brain/rpmbuild/BUILDROOT/scipy-0.16.1-6.fc24.x86_64/usr/lib64/python2.7/site-packages/scipy/linalg/_flapack.so: undefined symbol: sgegv_ The same for py3. I tried to rebuild scipy, got the same results.
Created attachment 1099888 [details] [PATCH] build deprecated functions also (RHBZ #1286349)
/usr/bin/strip:/home/brain/rpmbuild/BUILDROOT/lapack-3.6.0-3.fc24.x86_64/usr/lib64/liblapack_pic.a(sgegv.f): Unable to recognise the format of file: File format not recognized I have no idea.
Created attachment 1099922 [details] [PATCH] build deprecated functions also (RHBZ #1286349)
Lat patch fixes problem.
I can confirm that applying this patch to lapack, rebuilding lapack, and rebuilding atlas with the new lapack, fixes the problem with scipy. I have no idea whether this is the right solution, and who is at fault: scipy for using deprecated routines, or lapack for removing them. I couldn't find any docs about the deprecation.
(In reply to Zbigniew Jędrzejewski-Szmek from comment #5) > I can confirm that applying this patch to lapack, rebuilding lapack, and > rebuilding atlas with the new lapack, fixes the problem with scipy. > > I have no idea whether this is the right solution, and who is at fault: > scipy for using deprecated routines, or lapack for removing them. I couldn't > find any docs about the deprecation. http://www.netlib.org/lapack/lapack-3.6.0.html [Philippe Theveny, UC Denver] Deprecated routines CGEGS CGEGV CGELSX CGEQPF CGGSVD CGGSVP CLAHRD CLATZM CTZRQF DGEGS DGEGV DGELSX DGEQPF DGGSVD DGGSVP DLAHRD DLATZM DTZRQF SGEGS SGEGV SGELSX SGEQPF SGGSVD SGGSVP SLAHRD SLATZM STZRQF ZGEGS ZGEGV ZGELSX ZGEQPF ZGGSVD ZGGSVP ZLAHRD ZLATZM ZTZRQF matrix_order → matrix_layout in the LAPACKE C Interface. But in upstream's CMakeLists.txt I can see them as if(BUILD_DEPRECATED)
scipy has a bug open: https://github.com/scipy/scipy/issues/5266.
from my POV now we should build deprecated funcitons and in the future we will see what will happen in scipy.
So, scipy is slated to remove the wrappers for the deprecated functions in the next release. In the light of this, I think we should build lapack and atlas with the deprecated routines, to allow scipy to be unchanged. After scipy is released, drop the routines from lapack and atlas. If there are any other users in the distro, also wait a bit until they are adjusted.
next scipy release → scipy-0.17.0 to be precise.
(In reply to Zbigniew Jędrzejewski-Szmek from comment #9) > So, scipy is slated to remove the wrappers for the deprecated functions in > the next release. In the light of this, I think we should build lapack and > atlas with the deprecated routines, to allow scipy to be unchanged. After > scipy is released, drop the routines from lapack and atlas. If there are any > other users in the distro, also wait a bit until they are adjusted. Fully agree, but scipy needs rebuild anyway. But in this case we will not change its code.
can you do this? I dont have commit ACLs to lapack/scipy
Yes. I'll rebuild them in a few hours if nobody objects. Too many things depend on scipy, it would be really annoying to have it broken.
(In reply to Zbigniew Jędrzejewski-Szmek from comment #13) > Yes. I'll rebuild them in a few hours if nobody objects. Too many things > depend on scipy, it would be really annoying to have it broken. (02:53:05 AM) spot: ignatenkobrain: there is a semantic issue with your patch, but its harmless. go ahead and have zbyszek rebuild it. I'd do it, but I'm rather sick at the moment.
/me too Just getting ready, (but one version behind, as 6.9 is already release for some time, and may still take some time) to update sagemath to a newer version, as it is quite outdated now... Fails during documentation rebuild: [matroids ] /home/pcpa/rpmbuild/BUILD/sage-6.8/src/doc/en/reference/matroids/sage/matroids/matroids_plot_helpers.rst:11: WARNING: autodoc can't import/find module 'sage.matroids.matroids_plot_helpers', it reported error: "/usr/lib64/python2.7/site-packages/scipy/linalg/_flapack.so: undefined symbol: sgegv_", please check your spelling and sys.path
This was rebuilt today, the build for atlas finished at 6:47 UTC. Just make sure that you have that version. I'll close this, because it should be fixed for now. When new scipy comes out, this patch should probably reverted.