Tags: daneecool/AliceO2
Tags
CalDet<TPCPadFlags>: Protect against invalid reads A continuation of the CalDet<TPCFlags> saga, possibly related to https://its.cern.ch/jira/browse/O2-4671 Tests on ARM, even after deployment of the custom streamer in AliceO2Group#14830, still showed segfaults in TPC digitization. With the relevant code isolated into a unit test in AliceO2Group#14850, it was possible to do a valgrind study. This showed Invalid reads to the mData of CalArray. Thereafter, putting assert statements showed that we often access CalArray<PadFlags> data slightly out of bounds - irrespective of custom streamer or not. This then either indicates a problem in the code logic or a problem with the calibration CCDB objects. This should clearly be fixed. In the meantime, this commit adds a protection against invalid accesses and returns a trivial answer as well as an error message. This is in any case better than undefined behaviour. In addition, this commit introduces possibility to switch off the custom streamer for further studies.
CalDet<TPCPadFlags>: Protect against invalid reads A continuation of the CalDet<TPCFlags> saga, possibly related to https://its.cern.ch/jira/browse/O2-4671 Tests on ARM, even after deployment of the custom streamer in AliceO2Group#14830, still showed segfaults in TPC digitization. With the relevant code isolated into a unit test in AliceO2Group#14850, it was possible to do a valgrind study. This showed Invalid reads to the mData of CalArray. Thereafter, putting assert statements showed that we often access CalArray<PadFlags> data slightly out of bounds - irrespective of custom streamer or not. This then either indicates a problem in the code logic or a problem with the calibration CCDB objects. This should clearly be fixed. In the meantime, this commit adds a protection against invalid accesses and returns a trivial answer as well as an error message. This is in any case better than undefined behaviour. In addition, this commit introduces possibility to switch off the custom streamer for further studies.
Exercise CalArray<PadFlags> loading in a TPC unit test This makes it easy to valgrind the process. It seems that despite the custom streamer, there are still sporadic segfaults when exercising DeadChannelMapCreator::loadIDCPadFlags (at least on ARM64).
Exercise CalArray<PadFlags> loading in a TPC unit test This makes it easy to valgrind the process. It seems that despite the custom streamer, there are still sporadic segfaults when exercising DeadChannelMapCreator::loadIDCPadFlags (at least on ARM64).
Stability improvement for o2-sim startup Fixing a zeromq communication problem when the O2PrimaryServerDevice is too fast and quitting before O2HitMerge is even initialized. This can happen rarely, for instance when startup needs long due to CVMFS latency. This situation is now avoided by waiting for the permission to shutdown. The permission is given by O2HitMerger once the system is up and running (first actual hits have been received). Some minor cleanup (comment removal) in addition.
Stability improvement for o2-sim startup Fixing a zeromq communication problem when the O2PrimaryServerDevice is too fast and quitting before O2HitMerge is even initialized. This can happen rarely, for instance when startup needs long due to CVMFS latency. This situation is now avoided by waiting for the permission to shutdown. The permission is given by O2HitMerger once the system is up and running (first actual hits have been received). Some minor cleanup (comment removal) in addition.
Set proper EOR to anchored MC GRPECS (AliceO2Group#14843)
PreviousNext