Skip to content

HltEventLoopMgr: Print stack trace on the first soft timeout

Rafal Bielski requested to merge rbielski/athena:timeout-trace into master

The first time we flag a soft timeout from the timer thread of HltEventLoopMgr, call gSystem->StackTrace() to print the trace of all threads. Do this only once per process to avoid concurrent trace generation and having multiple traces in the same log file.

This is to help investigating very rare timeouts which cannot be reproduced offline, like ATR-25365.

Extra changes to ensure the timeout ART test still works correctly:

  • TrigP1Test: add the traces to check_log ignore patterns for the timeout test
  • TrigValTools: update check_log.py to avoid collecting traces on ignored lines
Edited by Rafal Bielski

Merge request reports