Closed maxnoe closed 1 month ago
Project ID: cta-observatory_ctapipe_AY52EYhuvuGcMFidNyUs
Can you think of a test that could catch this problem?
Telescopes being silently dropped from the trigger is quite serious, and worrying it was not caught for two years.
This condition only happens with the new configuration of Prod 6, before all 4 LSTs were identical.
We haven't really worked on anything Prod 6 yet, it's not even finished simulating.
I'd prefer to fixing this the way I do it here as the other change would require us to do a 0.22 and adapt configuration files.
With the fix at it is now, we can just publish an 0.21.1 without adapting configuration files.
I will add a test that tests for the bug.
Project ID: cta-observatory_ctapipe_AY52EYhuvuGcMFidNyUs
Rebased and changelog added, please approve again @Tobychev
Project ID: cta-observatory_ctapipe_AY52EYhuvuGcMFidNyUs
This was one of the weirder bugs I have ever tracked down. The symptom was that in consecutive runs of
ctapipe-process
on the same input file, sometimes LST-1 was missing from the output completely.Tracking it down to the SoftwareTrigger class, which used a mixture of using
TelescopeDescription
and its string representation for storing and looking up things.The dict
_ids_by_type
was the culprit, its content depended on the hash of str(TelescopeDescription), which was random, since hash of strings are random for security reasons in python.