Repository navigation
Conversation
There was a problem hiding this comment.
All reported issues were addressed across 2 files
Reply with feedback, questions, or to request a fix.
View guided diff | Re-trigger cubic
There was a problem hiding this comment.
All reported issues were addressed across 3 files (changes from recent commits).
Reply with feedback, questions, or to request a fix.
View guided diff | Re-trigger cubic
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #38.
A late dispatch still raises
QueueShutDown, but_start()rejects the shut-down queue before creating another run loop. Shutdown and cancellation reach the loop's existing exit handlers, and the temporary queue-get task is always cleaned up.Idle waits return after stop without restarting the bus, including waits already in progress. They use the existing idle signal, which the run loop sets after draining work and
stop()sets when aborting it. The queue and event state are checked again before returning while the bus is running.Reproduced the report's stop/dispatch sequence on Windows with Python 3.12: the original process stayed running after cancellation and used about one CPU core; the fixed version stays stopped and exits normally. Also reproduced a pending idle wait hanging during immediate stop on both the unchanged upstream base and the earlier PR head; it now returns.
Validation: nine shutdown/idle-wait cases pass, including a normal wait that must not return before its handler finishes. The full suite passes (147 tests, one existing duplicate-handler warning), as do the repository's pre-commit checks. The initial four shutdown regressions failed on the unchanged upstream base; two concurrent-stop regression cases failed on the previous PR head.