-
-
Notifications
You must be signed in to change notification settings - Fork 645
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Flaky/failing Windows integration tests #3431
Comments
Hi @vemv, preliminary analysis shows that in the failing case, that just before the nrepl-client is in between of closing the two session clones with the nrepl-server, it issues a expected sequence closes the two sessions in ids 9/10:
failing sequence which also tries to close sessions in 9/10, but it issues a classpath op in 11, before 10 has a chance to get the response back. It then is stuck in an endless bouncing until the test timeouts. The error is clearly indicating with the server responding that the session is unknown (
thanks |
Hi @ikappaki , thanks for the great analysis! I am able to see things more clearly now. The Lines 1771 to 1780 in 73f0d05
(a bit of context: the I wonder if we could avoid sending those requests if we detected that the session was Cheers - V |
Hi @vemv,
Yes, perhaps we could put something in the the other scenario is that the server never responded to the closing of the second session (the "tooling" session), and the cider nrepl client carried on witth its normal job? request to close the connection Lines 197 to 227 in 73f0d05
the close sync req (sending the two closing reqs, the one the server never response I believe is the Lines 1041 to 1044 in 73f0d05
The sync request, which will time out if no response comes back Lines 935 to 980 in 73f0d05
and we see that the close request has indeed timed out in the log
|
This was fixed with #3435 Again, thanks for the analysis, which lead me to a useful direction. Perhaps there are things that can be improved, namely:
Feel free to create an issue for that. |
Over the last few days, our Windows integration tests have tended to look like this:
...apparently it's always a single integration test that fails.
@ikappaki , would you like to have a look?
Thanks - V
The text was updated successfully, but these errors were encountered: