1 |
> EL Wed, 1 Jun 2011 06:57:03 +1200 |
2 |
> |
3 |
> Francois Bissey <fbissey@××××××××××××.nz> escrigué: |
4 |
> > > Hi, |
5 |
> > > |
6 |
> > > While updating sage and python to 4.7 and 2.7, I hit a variation of |
7 |
> > > an old bug (see http://bugs.gentoo.org/195619): scipy compiles with |
8 |
> > > lapack-reference but is unusable. Running in python |
9 |
> > > |
10 |
> > > import scipy.optimize |
11 |
> > > |
12 |
> > > fails with |
13 |
> > > |
14 |
> > > [...] |
15 |
> > > |
16 |
> > > from scipy.linalg import clapack |
17 |
> > > |
18 |
> > > ImportError: |
19 |
> > > /usr/lib64/python2.7/site-packages/scipy/linalg/clapack.so: undefined |
20 |
> > > symbol: clapack_sgesv |
21 |
> > > |
22 |
> > > I tried rebuilding numpy and scipy with lapack provided (eselect) by |
23 |
> > > lapack-reference, blas provided by blas-reference, cblas provided by |
24 |
> > > gsl and with/without clapack with the same result. |
25 |
> > > |
26 |
> > > Then I tried rebuilding numpy and scipy with lapack, blas and cblas |
27 |
> > > provided (eselect) by atlas and scipy worked flawlessly. |
28 |
> > > |
29 |
> > > Should scipy depend on a specific lapack provider instead of a |
30 |
> > > virtual? Enable it at compile time? Is it an unfortunate |
31 |
> > > combination what fails? |
32 |
> > > |
33 |
> > > I can provide build logs if you need them. |
34 |
> > |
35 |
> > Hi Marc, |
36 |
> > |
37 |
> > Just one thing that I can think of on the top of my head. We recently |
38 |
> > made scipy depend on arpack and superlu. They also have dependencies |
39 |
> > on blas/lapack if you didn't rebuild them as well they may cause |
40 |
> > trouble. |
41 |
> > |
42 |
> > I had a similar problem recently but eselecting everything properly |
43 |
> > worked for me. A last piece of info you could provide: are you using |
44 |
> > the main tree implementation or bicatali's overlay implementation? |
45 |
> > |
46 |
> > Francois |
47 |
> |
48 |
> Hi Francois, |
49 |
> |
50 |
> I tried rebuilding arpack and superlu without luck. Since you mentioned |
51 |
> that eselect fixed your problems, I also tried installing |
52 |
> cblas-reference and using it instead of gsl, also without luck. |
53 |
> |
54 |
> I'm using the main tree implementation. |
55 |
> |
56 |
Hi Marc, |
57 |
|
58 |
I have been using bicatali's overlay for a bit and I recompiled all the |
59 |
cblas/lapack stack from sage a few times with different implementations |
60 |
in the last few days. The only time i had an issue was when the value |
61 |
eselected was not what I wanted by accident. |
62 |
|
63 |
I am curious to see your build log for numpy and scipy. Also what does |
64 |
ldd /usr/lib64/python2.7/site-packages/scipy/linalg/clapack.so |
65 |
reports? |
66 |
|
67 |
Francois |