TAO test error [petsc-maint]

[Thomas Hisch]

The petsc4py unit tests fail with the following output:

======================================================================
ERROR: testOptions (test_object.TestObjectTAO)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test/test_object.py", line 62, in testOptions
    self.obj.setFromOptions()
  File "PETSc/TAO.pyx", line 124, in petsc4py.PETSc.TAO.setFromOptions (src/petsc4py.PETSc.c:185164)
Error: error code 86
[0] TaoSetFromOptions() line 402 in /home/jenkins/jobs/petsc-maint-complex/workspace/petsc/src/tao/interface/taosolver.c
[0] TaoSetType() line 2140 in /home/jenkins/jobs/petsc-maint-complex/workspace/petsc/src/tao/interface/taosolver.c
[0] Unknown type. Check for miss-spelling or missing package: http://www.mcs.anl.gov/petsc/documentation/installation.html#external
[0] Unable to find requested Tao type lmvm

[Lisandro Dalcin]

Is this with current petsc/maint and petsc4py/maint? I cannot reproduce.

[Thomas Hisch]

Yes this is with the current maint branches.

This are the flags used to compile petsc:

--download-scalapack
--download-suitesparse
--download-metis
--download-parmetis
--download-mumps
--download-scotch
--download-ptscotch
--with-c++-support=1
--with-scalar-type=complex
--with-x11=0
--with-shared-libraries=1
--with-fortran-kernels=1
--download-sowing
--with-c2html=0
--with-debugging=0

Did you try to test it with complex petsc scalars? It seems that I only get this test error with complex petsc scalars, but I have to double check it.

[Thomas Hisch]

Indeed, the error only occurs in a complex petsc scalar build.

[Lisandro Dalcin]

Oh, now I see. Indeed, TAO does not support complex scalars. I'll push a fix disabling these tests for complex builds.

[Lisandro Dalcin]

Fixed in maint and master: 171732539ff419c4ce7bcafff886ba3233a7f3f5

[Thomas Hisch]

Thx. I've just started a new jenkins build :-)

[Thomas Hisch]

It still does not work. Please use np.iscomplexobj instead of np.iscomplex to test whether the scalar is complex or not.

In [3]: np.iscomplex(0j)
Out[3]: False

[Lisandro Dalcin]

Sorry! Fixed: d6b990f1a74ae7fd002432131f53d56a238074fb

[Thomas Hisch]

thx!