ci: `sensors.tracing.TestKprobeMatchParentBinaries` might be flaky

[mtardy]

Potential flake:

https://github.com/cilium/tetragon/actions/runs/22990014669/job/66756456304?pr=4749

❌	pkg.sensors.tracing.TestKprobeMatchParentBinaries										(total:13 failed:2 skipped:0)	54.41798824s	13m48.10409944s

[kkourt]

CC: @kobrineli

[kobrineli]

Hmm. I'll check what's going wrong, thanks. Never met such problems while pushing commits to the other PR after merging parent binaries, though.

[kobrineli]

Did parent binaries tests failed in any other pr apart from the mentioned one?

[mtardy]

Hmm. I'll check what's going wrong, thanks. Never met such problems while pushing commits to the other PR after merging parent binaries, though.

Did parent binaries tests failed in any other pr apart from the mentioned one?

it's probably a flake i.e. something that might occur rarely because of a race/bug in the test, that's why if you can take a look and try to see what could have gone wrong that would be awesome so that we don't see it in the future. It's seems pretty rare though as I don't think I've seen it elsewhere so far.

[kobrineli]

Checked matchParentBinaries code and tests code for race conditions, haven't found the cause of the failure yet. Checked other CI jobs and CI at main branch, everything there is fine, so seems like a real rare case. Working on it

[dwindsor]

also being hit in #4817 :

Running test pkg.sensors.tracing.TestKprobeMatchParentBinaries [ 1139.501897] pkg.sensors.tra[8843] is installing a program with bpf_probe_write_user helper that may corrupt user memory!
........> failed after 51.806988311s: exit status 1

only on 5.10

[kobrineli]

I'll take a look