Open ikalash opened 1 week ago
reblessing seems okay to me if @bartgol doesn't see anything obvious in his PR.
The PR that removed warnings has nothing that looks suspicious. And yet, the tests started to fail right after that PR was merged. I'm trying to bisect on that PR, to see if I'm missing something. The diffs are indeed minor, but the fact that nothing should have changed bothers me.
Maybe it was a Trilinos change? old: 7b63078f259 new: 83e06e659ac
the only thing I see that might be relevant is the stk snapshot
I just ran git bisect, so I kept trilinos fixed to whatever I have on my laptop, and it pin pointed this commit:
b7ebca3a86bbdfe48987be48ed81b6afe058a38c is the first bad commit
commit b7ebca3a86bbdfe48987be48ed81b6afe058a38c
Author: Luca Bertagna <lbertag@sandia.gov>
Date: Mon Oct 7 11:51:12 2024 -0600
Fix misc warnings borderline buggy
src/disc/omegah/Albany_OmegahMeshFieldAccessor.cpp | 2 ++
src/evaluators/gather/PHAL_GatherSolution_Def.hpp | 6 ++++--
src/landIce/evaluators/LandIce_SimpleOperation.hpp | 2 +-
tests/unit/disc/generic/DummyConnManager.cpp | 6 +++---
4 files changed, 10 insertions(+), 6 deletions(-)
I thought none of these changes mattered for FOThickness, but it's not true:
So we were actually using H0 (the initial thickness) for other parts of the calculations (in the velocity solver). This is an example of why warnings are not to be taken lightly...
I think we can bless and close this issue.
Awesome. Thanks all for the sleuthing. And for fixing the BinarySum op.
This test began a couple weeks ago with diffs:
https://sems-cdash-son.sandia.gov/cdash//test/8919018?graph=status
The diffs are tiny actually:
Presumably these are due to a PR Luca pushed a couple weeks ago.
The diffs are very small. Should we just rebaseline?