Repository navigation
http2: submit RST_STREAM before emitting 'aborted' - #66314
Merged
nodejs-github-bot merged 1 commit intoSep 29, 2026
Merged
Conversation
Signed-off-by: lazerg <lazerg2@gmail.com>
Collaborator
|
Review requested:
|
nodejs-github-bot
added
http2
Contributor
|
Caution AgentScan found account activity patterns that may be consistent with automation. This is a heuristic, not proof that this pull request was opened by an agent or violates policy. AI-assisted contributions are permitted, but automated tooling must not open pull requests without advance approval, and contributors must personally understand, test, verify, and take responsibility for every submitted change. See the AgentScan analysis, AI use policy, and automation policy for additional context. |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #66314 +/- ##
==========================================
+ Coverage 90.36% 90.37% +0.01%
==========================================
Files 790 790
Lines 274273 274498 +225
Branches 52513 52569 +56
==========================================
+ Hits 247842 248085 +243
+ Misses 16897 16880 -17
+ Partials 9534 9533 -1
🚀 New features to boost your workflow:
|
panva
added
author ready
github-actions
Bot
removed
the
request-ci
Collaborator
pimterry
approved these changes
Sep 29, 2026
pimterry
added
the
commit-queue
Collaborator
|
Landed in 1f26576 |
nodejs-github-bot
removed
the
commit-queue
aduh95
added
the
backport-requested-v26.x
Contributor
|
This doesn't land cleanly on v26.x-staging, and would require a manual backport if we want it there |
robobun
added a commit
to oven-sh/bun
that referenced
this pull request
Oct 10, 2026
The first commit stores the code of close(code) before its listeners run. A review and fuzzing found rows that changed against main for callers without the bug. This commit puts them back, and follows node main where a listener throws. - close() keeps its late store of the code too. Another closer that a listener runs (a session teardown, the AbortSignal of the request) does not replace the code at 'close'. - A listener that throws out of close() does not stop the close. The writable ends, the reset goes out and 'close' fires, as in node main (nodejs/node#66314). The error is thrown again after that. - A close() from a listener of a running close() validates its arguments and keeps its callback. Only its code is ignored. - When a listener destroyed the stream, _destroy has queued the reset and close() queues no second one. The reset of close(0) at 'finish' stays. - While close() runs, the stored code does not count as a reset that happened: the streamEnd handlers wait for 'end', and a client gets the response that arrives inside close() on an in-process transport. - _destroy does not keep NGHTTP2_NO_ERROR for a pushed stream, and the teardown of a server session does not replace a stored code.
robobun
added a commit
to oven-sh/bun
that referenced
this pull request
Oct 10, 2026
The first commit stores the code of close(code) before its listeners run. A review and fuzzing found rows that changed against main for callers without the bug. This commit puts them back, and follows node main where a listener throws. - close() keeps its late store of the code too. Another closer that a listener runs (a session teardown, the AbortSignal of the request) does not replace the code at 'close'. - A listener that throws out of close() does not stop the close. The writable ends, the reset goes out and 'close' fires, as in node main (nodejs/node#66314). The error is thrown again after that. - A close() from a listener of a running close() validates its arguments and keeps its callback. Only its code is ignored. - When a listener destroyed the stream, _destroy has queued the reset and close() queues no second one. The reset of close(0) at 'finish' stays. - While close() runs, the stored code does not count as a reset that happened: the streamEnd handlers wait for 'end', and a client gets the response that arrives inside close() on an in-process transport. - _destroy does not keep NGHTTP2_NO_ERROR for a pushed stream, and the teardown of a server session does not replace a stored code.
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
6 participants
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.
closeStream() emitted 'aborted' before submitting RST_STREAM. If an 'aborted' listener destroyed the stream (stream.pipeline() does this on its own), _destroy() skipped its RST because the stream was already marked closed, then destroyed the handle, so the later submitRstStream() found no handle and nothing was sent. The peer's stream stayed open until the session closed.
This emits 'aborted' after the RST_STREAM has been submitted.
Fixes: #66306