Let's investigate why... it (again) seems to be only the "file conflicts" check. The timeout for the test is set to 90 minutes per architecture and it is still getting timeout-killed from time to time.
big packages = packages with lot of files.
@msrb is there anything I can help with to investigate?
@mvadkert not really, first I am going to enable debug logs (redirected to a file) so I can see what was happening there when the job was timeout-killed.
But thanks ;)
I think we are hitting this on a recent llvm update:
https://osci-jenkins-1.ci.fedoraproject.org/job/fedora-ci/job/rpmdeplint-pipeline/job/master/34656/
Would it be possible to split out the rpmdeplint sub-tests? I am mostly interested in running the check-repoclosure test.
Finally we think we found the problem and are investigating how to solve it. We believe we will make it possible to make the testing ~3x times faster so for these large components we should see a big gain, from 10+ hours to 2-3 hours.
This is tracked in this RH Jira for now: https://issues.redhat.com/browse/OSCI-3983
@msrb would you have time look at rpmdeplint if it could not have similar problems as rpminspect, going outside of the PWD, and getting bad performance because of this?
@mvadkert I will take a look. rpmdeplint is not IO heavy. It downloads a lot of stuff, but there were some improvements that landed upstream.
let me try llvm on a pristine machine
Nah, my machine is gone. But @msrb promised to look at rpmdeplint
Sounds like RHBZ#1465736 for which the fix was released in 2.0rc3
Log in to comment on this ticket.