toolbox testsuite fails on debian bookworm

[prudhomm]

The toolbox testsuite fails on debian bookworm here :

• https://buildkite.com/feelpp/feelpp/builds/4969#01893ae0-85de-402b-92c6-ef16dd6d4f3a/3020-4747
  it is the NAFEMS LE10 benchmark which fails
• https://buildkite.com/feelpp/feelpp/builds/4969#01893ae0-85de-402b-92c6-ef16dd6d4f3a/3020-5178

More precisely the solver is configured to use gamg and it fails for debian:12 using petsc 3.18.
Note that it worked in previous version, there is a chance that it is a PETSc bug.
Changing to pc-type=lu fixes the testcase.

[prudhomm]

@vincentchabannes I discussed with @prj- , he confirmed that in PETSc 3.18 there is internal changes in gamg that could broke the solver. @prj- will have a look at a mail discussion that discussed these issues. It seems that we were not the only ones hit by this. there are also changes between 3.8 and 3.19 that we will need also to tackle

[prj-]

(I'm still looking for the proper PETSc thread 😓)

[prj-]

There we go! https://lists.mcs.anl.gov/pipermail/petsc-users/2023-April/048337.html

[prudhomm]

that's interesting because it is exactly the same benchmark we use ! NAFEMS LE10

[prudhomm]

here is an interesting answer regarding how to play with the new options
https://lists.mcs.anl.gov/pipermail/petsc-users/2023-April/048366.html

[prj-]

that's interesting because it is exactly the same benchmark we use !

That's an uncanny coincidence indeed.