Antonino Ingargiola wrote:
2009/7/20 Robert Cimrman cimr...@ntc.zcu.cz:
Antonino Ingargiola wrote: <snip>
Recompiling sfepy now I have only one error in tests/test_input_acoustics.py. Here it is the output: I guess you needed no recompilation in this case - sfepy uses umfpack only as an external package with no link-time dependence.
So, if I understand correctly I don't need the UMFPACK headers either? Maybe just installing the umfpack scikit with the runtime UMFPACK library should be sufficient, right? I ask this because the runtime library is automatically installed by scipy (on ubuntu) while the -dev package is not... and I'm taking notes on the correct procedure on ubuntu 9.04. At the end I can post the instructions somewhere.
You need the headers (umfpack.h) to build the umfpack scikit (or the umfpack wrappers within scipy). You do not need the headers to build sfepy. So my point was that after installing umfpack, sfepy does not need to be recompiled.
It would be great if you post the installation instructions at [1], thanks!
[1] http://code.google.com/p/sfepy/wiki/Installation
$ ./runTests.py --debug tests/test_input_acoustics.py <<< tests/test_input_acoustics.py sfepy: nls: iter: 0, residual: 1.116698e+07 (rel: 1.000000e+00) warning: singular matrix zeroing nan and inf entries... warning: (almost) singular matrix! (estimated cond. number: inf) sfepy: rezidual: 0.00 [s] sfepy: solve: 0.00 [s] sfepy: matrix: 0.00 [s] sfepy: linear system not solved! (err = 1.116698e+07) sfepy: nls: iter: 1, residual: 1.116698e+07 (rel: 1.000000e+00) ... ../input/acoustics.py solved --- test_input: failed! !!! 1 test failed 1 test file executed in 0.51 s, 1 failure(s) of 1 test(s) This is strange, look at what I get:
sfepy: nls: iter: 0, residual: 1.116698e+07 (rel: 1.000000e+00) sfepy: rezidual: 0.00 [s] sfepy: solve: 0.00 [s] sfepy: matrix: 0.00 [s] sfepy: nls: iter: 1, residual: 8.460917e-08 (rel: 7.576731e-15) ... ../input/acoustics.py solved
- no singular matrix warning. I have tested with numpy 1.2.1, scipy 0.7.0, umfpack 5.2.0 on gentoo, latest git version of sfepy. But I do not use the scikit (good old deprecated scipy.sparse.linalg.dsolve.umfpack). Let me check with the latest scikit
- yes, it works too.
I've tested with the numpy 1.2.1 and scipy 0.7.0 shipped with ubuntu and I get the same error (see my first mail).
What is you swig version?
I use the swig version in ubuntu 9.04 that is the 1.3.36.
I have 1.3.39. I will try to downgrade and see.
r.