[petsc-dev] Regression tests

[Jed Brown]

On Thu, Jul 26, 2012 at 11:08 AM, Satish Balay <balay at mcs.anl.gov> wrote:

> I think python is fine - as long as the run targets still remain in
> the makefile.
>
> Are you saying each 'run' target in the makefile will be tagged - and
> the python script will parse this tag?
>

I meant getting them out of the makefile so we can declare the test in
Python. An example syntax would be

test.py:
from Petsc.Test import InModule, Exec, Test, Main

with Exec(sources='ex19.c'): # ex19 inferred from name, could specify
explicitly
  Test('fas', np=8, args='-snes_type fas -snes_monitor_draw', requires='X')

if __name__ == '__main__': Main()

The Test() declaration would register itself as being associated with
ex19.c. We can add keywords to create batches. The output file is inferred
from the test name, so here we would compare with output/ex19_fas.out. This
should allow

./test.py build ex19

./test.py check ex19_fas

./test.py check ex19_\* # run all the ex19 tests

./test.py check -n 16 --if '(X || Fortran) && !Complex'

The last line would mean to run all tests requiring X or Fortran, but not
complex, using up to 16 MPI processes total.

We could preserve "make runex19_fas" in the transition period, but I'd like
to be able to do everything from test.py.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120726/e0c2dee8/attachment.html>