scautopick stop working after ~3 hours about main HOT 6 CLOSED

Hi Luca,
good catch! 200k * 3600 * 3 is very close to 2**31 so you are probably on the right track that some counter is the culprit.

find ~/seiscomp/include/seiscomp -name "*.h"|xargs grep sampleCount

Take your pick. :) Maybe

https://github.com/SeisComP/common/blob/master/libs/seiscomp/math/filter/stalta.cpp#L90
https://github.com/SeisComP/common/blob/master/libs/seiscomp/math/filter/stalta.cpp#L97

as a start... The developer on duty certainly did not have such large _sampleCount on his radar, let alone the possibility that it may become negative... I am pretty sure that changing the type of _sampleCount to int64_t will already fix the issue. Would you mind trying that out?

Thanks for the hint!
Joachim

from main.

I would just skip increasing the samplecount once it is not needed anymore, after it is >= numLta. If a lta time window for 1Mhz data of 30 minutes should be supported then we can use int64 anyway. Changing the count possibly breaks binary compatibility.

from main.

Thank you, I had the feeling that you knew something already :) I will test it and report back.

from main.

I confirm that is the bug

from main.

I have fixed it hopefully, please give it another try. The fix is in common repo.

from main.

Thank you!

from main.