Hi, On 26-12-2022 20:58, Laurent Bigonville wrote:
> I looked at the results of the autopkgtest of your package. I noticed > that it regularly fails (after the fix for bug #1019221).
It's impossible for me to reproduce this locally, can you still see that in the CI runs?
You can trigger the test yourself to. Just hit one of the recycle (♲) buttons (I recommend using a migration-reference/0 run for those retries).
For me this looks like a race condition,
That's a very common reason for tests to be flaky.
the test code already waits for 2 sec after the start of the nut daemon an drivers to be certain that everything is properly started. I can increase that sleep, but that's probably not the best solution.
Maybe not the best, but definitely easy. And I suspect good enough if you make it big enough. Wasting a couple of seconds on this in CI tests is not a bad thing.
An other solution is to have a loop checking that the (dummy) driver is really connected, but I need to investigate how and I've no time ATM.
Unless the test remains flaky if you increase the timeout to something like 20 or 30 seconds (if the daemon normally returns within a second) I wouldn't spend too much time on it if I were you. Because also such a loop would need a timeout to prevent it from hanging in case the daemon never returns.
Last solution could be to ask the release team to tag this bug as "bookworm-ignore" so it can be part of the upcoming release and try to fix this for the next one.
I would rather expect that the bug is then "fixed" by disabling the test. Flaky tests are really resource draining for people that are not related to the package, so I'm absolutely not in favor of bookworm-ignoring them.
WDYT?
Just raise the timeout (considerably) and be done. Paul
OpenPGP_signature
Description: OpenPGP digital signature
