Modify with storing the keys to erase and remove them outside the for loop that contains the map.
On tests, I see the same number of triggers and files that crashed don't crash anymore.
Joint here are two valgrind files running on the same files to make CAFs. I'm not a valgrind expert but the ==4445== Invalid read of size 8 that seem to be associated to buildTriggerList have disapeared in the postfix version, so I assume that fixed some memory issue somewhere?
logfile_valigrind_postfix.txtlogfile_valigrind_prefix.txt
Some Minirun4.5 jobs were failing at the trigger filling procedure, on test I've run, it seems that it happens even more with MINERvA only events.
As I understand it, the erase statement in the trigger matching for loop seem to remove an element we try to access to afterwards https://github.com/DUNE/ND_CAFMaker/blob/main/src/makeCAF.C#L278
Modify with storing the keys to erase and remove them outside the for loop that contains the map. On tests, I see the same number of triggers and files that crashed don't crash anymore.
Joint here are two valgrind files running on the same files to make CAFs. I'm not a valgrind expert but the
==4445== Invalid read of size 8that seem to be associated tobuildTriggerListhave disapeared in the postfix version, so I assume that fixed some memory issue somewhere? logfile_valigrind_postfix.txt logfile_valigrind_prefix.txt