| Summary: | sci-libs/scipy: missing dependency | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Michael Davidsaver <mdavidsaver> |
| Component: | New packages | Assignee: | Gentoo Science Related Packages <sci> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | ||
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
Added virtual/cblas as a dep, which should be enough. Thanks for catching this one. (In reply to comment #1) > Added virtual/cblas as a dep, which should be enough. Thanks for catching this > one. > I have cblas, blas and lapack installed with reference and atlas, but i still watching the error =S Some idea? Same problem here when I am using cblas-atlas instead of cblas-reference. (In reply to comment #2) > > I have cblas, blas and lapack installed with reference and atlas, but i still > watching the error =S (In reply to comment #3) > Same problem here when I am using cblas-atlas instead of cblas-reference. Make sure you have the eselect profiles set to atlas or atlas-threads for blas,cblas and lapack before emerging both numpy and scipy if you want to use atlas in scipy. If it still doesn't work, and not related with the subject of this bug, please open a new bug with your build log and eselect profiles. Thanks, |
The scipy-0.6.0-r4 ebuild assumes cblas is installed, but does not depend on virtual/cblas. I belive this is because dev-python/numpy does, but it is a conditional dependency (lapack use flag). When cblas is missing Scipy gets non-fatal errors from pkg-config during the build. While the build does complete what results is partially broken. I get an undefined symbol 'daxpy_' (part of blas) when trying to import 'scipy.integrate' and things which depend on it. When scipy is build with cblas present this import works correctly. Blas-reference was present at all times. It would seem that scipy needs to depend on virtual/cblas directly, or depend on numpy with the lapack flag. I suspect the second would be better, but am not sure. >>> import scipy.integrate Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/lib64/python2.5/site-packages/scipy/integrate/__init__.py", line 10, in <module> from odepack import * File "/usr/lib64/python2.5/site-packages/scipy/integrate/odepack.py", line 7,in <module> import _odepack ImportError: /usr/lib64/python2.5/site-packages/scipy/integrate/_odepack.so: undefined symbol: daxpy_