Dealii: petsc/solver_(01|02) and petsc_complex/solver_(01|02) tests broken.

[tamiko]

All 4 tests throw a non-convergence error.

Prior to commit 6f53784f (back in 2013) this was apparently undetected. I accidentally checked in wrong comparison files with 6f53784f.

[tamiko]

In reference to #2814

[tamiko]

One more tests that test for a single precision/maximal iteration count and simply fail:

petsc/solver_01
petsc/solver_02
petsc/solver_11
petsc_complex/solver_real_01
petsc_complex/solver_real_02

We have other tests (e.g. lac/solver) that have intermittent NoConvergence errors - but test a variety of solvers and solver control values.

[tamiko]

@davydden @oneliefleft I guess none of the three original petsc/solver_* tests were ever meant to simply fail.

[davydden]

@tamiko judging from

 // it needs to be expected for the
 // richardson solver that we end up
 // here, so don't abort

i would say they were meant to fail.

[tamiko]

@davydden _ugh_ This doesn't make much sense. And funny enough it was me who wrote that.

[tamiko]

(the comment, not the tests)

[tamiko]

@davydden I'll fix it. It turns out that we reach a convergence after around 1800 iterations.