Skip to content

GitLab

Add --ignore-broken-inputs option to keep running HltEfficiencyChecker if encounter a dodgy file

Merged Ross John Hunter requested to merge rjhunter-sidestep-FileInputError into master

FYI @rmatev interested to hear your opinion on this.

As reported awhile ago by @mwhitehe (thanks!), when you run the HltEfficiencyChecker wizard, if you come across a file that cannot be opened for whatever reason, the trigger & tupling will skip it and keep running, making a perfectly valid tuple for analysis. ApplicationMgr will however exit with a not-very-descriptive error code 2, and so our check_call command in wizard.py will throw a CalledProcessError, and HltEfficiencyChecker will stop.

This is not ideal behaviour though:

  • You may consider the tuple you've made to be perfectly valid since the dodgy file was simply skipped,
  • The error code doesn't easily tell you what the problem was - the rest of the log shows the successful completion of a job,
  • Grid sites often go down - since the tupling & trigger will naturally just keep going, why can't HltEfficiencyChecker keep going.

I stop short of enforcing any behaviour, and print a more descriptive message if we encounter this. I added the --ignore-broken-inputs arg so the user can get HltEfficiencyChecker to keep going if that is what they wish.

Edited by Ross John Hunter