Matti Kortelainen
Matti Kortelainen
> Can I have a recipe with the build/config that reproduces it? Still on cmsdev42 ```sh cmsrel CMSSW_15_1_RNTUPLE_X_2025-07-06-2300 cd CMSSW_15_1_RNTUPLE_X_2025-07-06-2300/src cmsenv cmsRun /build/mkortela/debug/issue48400/CMSSW_15_1_RNTUPLE_X_2025-07-06-2300/src/reproducer_cfg.py ```
We could also rebuild the debug build with `RelWithDebInfo` instead of `Debug` if that would help.
> > sorry for the probably dumb question, but is a duplicated sequence an issue? > > If the config builder doesn't complain I wouldn't think it might cause problems....
assign RecoMuon/MuonIdentification
FYI @cms-sw/muon-pog-l2
The crash occurring in jemalloc hints towards a memory problem. I'd suggest to run the reproducer in ASAN or UBSAN or through valgrind.
The failures were last seen on CMSSW_15_1_X_2025-06-27-2300 where the jobs were ran on V100 (see e.g. [log](https://cmssdt.cern.ch/SDT/cgi-bin/logreader/el8_amd64_gcc12/CMSSW_15_1_GPU_X_2025-06-27-2300/pyRelValMatrixLogs/run/160.03502_HydjetQ_MinBias_5362GeV_2023_ppReco/step4_HydjetQ_MinBias_5362GeV_2023_ppReco.log#/)). All tests since then were ran on H100, so IB tests alone are...
Explicit tests on T4 sounds sufficient to me. Thanks!