[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#1006648: transition: numerical stack: slepc petsc hypre scotch scalapack



On 2022-03-16 09:27, Sebastian Ramacher wrote:
On 2022-03-15 20:33:02 +0100, Drew Parsons wrote:
Control: block 1006648 by 1006615

On 2022-03-05 12:29, Sebastian Ramacher wrote:
> Control: tags -1 confirmed
>
> On 2022-03-01 13:34:21 +0100, Drew Parsons wrote:
> >
> > The transitions are:
> >
> > scalapack  2.1 → 2.2
> > scotch     6.1 → 7.0
> > hypre      2.22 → 2.23
> > petsc      3.15 → 3.16
> > slepc      3.15 → 3.16

All transition packages uploaded and built, but got some turbulence right at
the end from the pybind11 clients (dolfin and dolfinx).

Some pybind11 patches are already available, will start applying.

Thanks for the quick pybind11 fix.

fenicsx-performance-tests failed with MPI errors on some architectures.


Not clear yet what's going on with fenicsx-performance-tests. A new release is expected soon, perhaps some of the updates will have fixed tests with PETSc 3.16.

fenicsx-performance-tests is at the end of the chain. At this point I recommend just removing fenicsx-performance-tests from testing so as not to hold up the transition while we figure out what's wrong with the performance tests. The dolfinx operational tests (fenics-dolfinx build and debci tests) are passing fine (except s390x which has always been failing debci, and test_sub_bbtree_box reported now at https://github.com/FEniCS/dolfinx/issues/2070). The test_sub_bbtree_box failure is unfortunate but sporadic. I don't think it's worth holding up the transition for it (only one dolfinx test failure in 312, can be skipped temporarily if needed).

I'm comfortable removing fenicsx-performance-tests from testing for now.

Drew


Reply to: