Facky tests revealed mainly with faster CI (aka Github Actions but also with Java 25)#1610
Conversation
jeanouii
force-pushed
the
fix/flacky-tests-github-actions
branch
2 times, most recently
from
1d892cb to
541942e
Compare
jeanouii
force-pushed
the
fix/flacky-tests-github-actions
branch
from
c2a5a80 to
a695119
Compare
…ailover During failover in transacted sessions with async dispatch (MessageListener), prefetched messages sitting in the unconsumedMessages buffer were not being tracked in previouslyDeliveredMessages. This caused them to be incorrectly identified as duplicates on redelivery and poison-acked to the DLQ.
Give the GC a bit more time to operate before failing Add some tolerance to JVM timing variations. The test is to validate the priority ordering not the overall performance, so it's ok Fix test stability Add some flexibility to the test assertion
Use Ephemeral ports when possible Fix race condition in test Improve stability Harden Stomp test because of timing issues/race conditions Move the JMSCOntext.start() a bit later to avoid race conditions
jeanouii
force-pushed
the
fix/flacky-tests-github-actions
branch
from
d3cb053 to
f4ebde6
Compare
jbonofre
left a comment
jbonofre
left a comment
There was a problem hiding this comment.
I reviewed all the updated test and all looks good, fixing several race conditions.
|
I don't understand why we are combining multiple unrelated changes in one PR? This doesn't make sense at all to combine several Jiras of changes into one commit. We need to be able to isolate issues by jira to rollback or fix |
|
The consumer changes to the client especially need to be reviewed more they are higher risk changes and should be by itself |
|
OK, the change doesn't look problematic to me. I propose to keep the test updates and revert the client change for broader discussion then. @jeanouii do you mind to create a PR to revert client change ? |
…s but also with Java 25) (apache#1610)" This reverts commit 94c3c3d. Note: Reverting due to accidental inclusion of [AMQ-9829] ActiveMQMessageConsumer.java
|
@jbonofre - Those changes themselves may or may not be fine, but they need to be done in isolation so that they can be noticed. I didn't even notice it was part of this because the title of the PR only mentions flaky tests and faster CI. Anything with a bug fix, especially I think needs to be separated out. Also, if it really is a bug we should backport the fix but we may not necessarily want to backport the rest of the commit. |
3 participants
Includes various fixes related to race conditions or bugs