Move owned WebSocket back to api::WebSocket on final hibernation event #664
Conversation
This file contains 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
MellowYarker
commented
May 18, 2023
MellowYarker
commented
May 18, 2023
|
(Just as another bit of cleanup I noticed, probably want to mark the destructors for |
xortive
approved these changes
May 18, 2023
jasnell
reviewed
May 18, 2023
| @@ -699,7 +699,7 @@ void WebSocket::ensurePumping(jsg::Lock& js) { | |||
| KJ_FAIL_ASSERT("Unexpected native web socket state", native.state); | |||
| } | |||
| } | |||
| }, [this](jsg::Lock& js, jsg::Value&& exception) mutable { | |||
| }, [this, thisHandle = JSG_THIS](jsg::Lock& js, jsg::Value&& exception) mutable { | |||
There was a problem hiding this comment.
As an alternative here, the JSG_THIS could be attached to the kj::Promise returned by the then(...) rather than captured by the lambda. Would save having to grab it twice. Obviously not a big deal tho.
There was a problem hiding this comment.
I suppose, longer term, we'd want these to maybe be JSG_VISITABLE_LAMBDA()s, so that the garbage collector had visibility into them? But that seems to be a preexisting issue.
MellowYarker
force-pushed
the
milan/hibernation-lifetime
branch
2 times, most recently
from
bba7237
to
100e65a
Compare
jclee
reviewed
May 18, 2023
jclee
approved these changes
May 18, 2023
We saw a situation where queued websocket messages (send()/close()) that were inserted during the webSocketClose event failed to deliver. This was caused by the fact that the HibernationManager drops HibernatableWebSocket's after their final close/error event is dispatched. Unfortunately, the HibernatableWebSocket would get dropped too soon, so the underlying kj::WebSocket was no longer available to send messages. We fix this by moving the owned kj::Own<kj::WebSocket> out from the HibernatableWebSocket and back into the api::WebSocket's NativeState after dispatching the final close/error event. NativeState's lifetime is managed by the IoContext's DeleteQueue, so it lives long enough to flush the remaining messages.
Related to the hibernation work; we noticed a few other things that could potentially lead to problems.
MellowYarker
force-pushed
the
milan/hibernation-lifetime
branch
from
5655c42
to
6a32783
Compare
|
Thanks for reviewing! |
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.
We saw a situation where queued websocket messages (
send()/close()) that were inserted during thewebSocketCloseevent failed to deliver. This was caused by the fact that theHibernationManagerdropsHibernatableWebSocket's after their final close/error event is dispatched.Unfortunately, the
HibernatableWebSocketwould get dropped too soon, so the underlyingkj::WebSocketwas no longer available to send messages.We fix this by moving the owned
kj::Own<kj::WebSocket>out from theHibernatableWebSocketand back into theapi::WebSocket'sNativeStateafter dispatching the final close/error event.NativeState's lifetime is managed by theIoContext'sDeleteQueue, so it lives long enough to flush the remaining messages.