It's hard to see at the moment (all red Dashboard), but I think this is related to the PR like Out-Of-Bounds accesses (OOBs):

At line 299 of file /opt/cp2k/src/qs_tddfpt2_bse_utils.F
Fortran runtime error: Index '12' of dimension 1 of array 'csvirt%local_data' above upper bound of 11
At line 298 of file /opt/cp2k/src/qs_tddfpt2_bse_utils.F
Fortran runtime error: Index '9' of dimension 1 of array 'wxvirtao%local_data' above upper bound of 8

@annahehn, could you please take a look at the failing regtest-tddfpt-bse/H2O.inp test.

The code somehow seems to mess up global/local indexes/bounds to some extent (this is only a very high-level first look). Perhaps it was not tested thoroughly with PSMP.

Yes, there a few places where %local_data is being accessed, which is always error prone. Ideally, high-level routines from the fm package would be called instead.

@annahehn is already working on this and I believe @fstein93 is helping. However, since it's now been over a month three weeks, we should maybe consider a rollback to recover the dashboard.

Well, the branch/code is not lost if reverted. It's however an inconvenience to reapply with a new PR. In any case, it's a new PR and I am leaning towards a rollback if there are no conflicts or additional work for the person who reverts it. The current CI or Dashboard state seems degenerated like no clear visibility anymore if current/later contributions are valid or not. That's fine, I think, if we had no contributions or upcoming release. Depending on the timeline (far release or resolution), I would leave it or not.

I am working on fixing the bugs.

I do have an idea on how to fix it but it will take a bit of time to mimic the intended tensor contraction. I am currently in exchange with @annahehn on how to proceed.