833462 - EventSource (SSE) throws JS error on page reload

Status: RESOLVED FIXED

(Core :: DOM: Networking, defect, P2)

Description

[Dana Basken]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_8_3) AppleWebKit/536.28.8 (KHTML, like Gecko) Version/6.0.3 Safari/536.28.8

Steps to reproduce:

Created an EventSource, verify it is connected, reload page.


Actual results:

Receive console error "The connection to http://<hostname>/<path>/eventsource.php was interrupted while the page was loading."


Expected results:

The EventSource should have closed silently.

Comment 1

[Matthias Versen [:Matti]]

Please provide a testcase/public URL.

Comment 2

[Dana Basken]

Attachment: Example server php script for this bug

This is a server php script to demonstrate this bug.

Comment 3

[Dana Basken]

Attachment: Example html file for this bug

Comment 4

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

Olli or Masatoshi-san, could you have a look at what's happening here?

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

We just print stuff to console as expected, but I guess we should try to not do that in this case.

Comment 6

[Dirkjan Ochtman (:djc)]

Can someone provide a few hints about how to fix this? It's pretty annoying for us, and I might be able to hack on it.

Comment 7

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

(In reply to Dirkjan Ochtman (:djc) from comment #6)
> Can someone provide a few hints about how to fix this? It's pretty annoying
> for us, and I might be able to hack on it.

You probably want to skip sending a message to the console if the result code EventSource::OnStopRequest gets is NS_BINDING_ABORTED.

Comment 8

[Dirkjan Ochtman (:djc)]

In particular, I'd also want to not fire the EventSource error event.

Comment 9

[Dana Basken]

Also, please note that as of (at least) FireFox 22.0, this same error is now showing up for WebSocket. Not sure if it's the same fix, but both EventSource and WebSocket should not throw a console error on page refresh. I think the request to also not fire the "error" event should be fixed as well.

Comment 10

[Espen Hovlandsdal]

Attachment: Bug 833462: Don't log EventSource error on NS_BINDING_ABORTED

Review commit: https://reviewboard.mozilla.org/r/206028/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/206028/

Comment 11

[xo]

For those that need it a work around I'm using is to close the eventSource before the page closes.

const eventSource = new EventSource('/api/v1/firehose');
window.addEventListener('beforeunload', () => {
	eventSource.close();
});

Issue is still present in Firefox 95.0.2 (64-bit)

Comment 12

[Fynn]

Without the

window.addEventListener("beforeunload", handleClose);

i get the following error:

Firefox can’t establish a connection to the server at http://localhost:3001/v1/stream/events.

when i add the "beforeunload", i still get an error, but a different one:

Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at http://localhost:3001/v1/stream/events. (Reason: CORS request did not succeed). Status code: (null).

Does anyone else have the same problem?

(This is not a general CORS problem. The connection is established without further problems. It is exclusively about the reload).

Comment 13

[Fynn]

(In reply to Fynn from comment #12)

Without the

window.addEventListener("beforeunload", handleClose);

i get the following error:

Firefox can’t establish a connection to the server at http://localhost:3001/v1/stream/events.

when i add the "beforeunload", i still get an error, but a different one:

Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at http://localhost:3001/v1/stream/events. (Reason: CORS request did not succeed). Status code: (null).

Does anyone else have the same problem?

(This is not a general CORS problem. The connection is established without further problems. It is exclusively about the reload).

Comment 14

[Hsin-Yi Tsai (she/her)[:hsinyi]]

Moving to Necko per the component description

Comment 15

[Sunil Mayya]

Attachment: Bug 833462 - suppress NS_BINDING_ABORTED errors for SSE. r=#necko

Comment 16

[Sunil Mayya]

We can suppress the error to the console when we are navigating away from the page or reloading.
This patch should resolve the  bug -> https://phabricator.services.mozilla.com/D295721

This bug falls in similar pattern to the other bugs we have - when a page navigates, the APIs (fetch, XHR, SEE) reporting error.

We need to suppress error logs for cases where we navigate away/ reload. The only problem is, it safe to assume NS_BINDING_ABORTED as user-initiated error (due to navigation, reloads)? We will discuss this internally and confirm.

Comment 17

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/0d35f4d1fd63
https://hg.mozilla.org/integration/autoland/rev/c85571dea524
suppress NS_BINDING_ABORTED errors for SSE. r=necko-reviewers,valentin

Comment 18

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/8c7a68a4c261
https://hg.mozilla.org/integration/autoland/rev/faf255707f0b
Revert "Bug 833462 - suppress NS_BINDING_ABORTED errors for SSE. r=necko-reviewers,valentin" for causing wpt failures at EventSource

Comment 19

[Silaghi Andreea]

Backed out for causing wpt failures at EventSource
Backout Link
Push with failures
Failure Log
Failure line TEST-UNEXPECTED-TIMEOUT | /eventsource/eventsource-request-cancellation.window.html | EventSource: request cancellation - Test timed out

Comment 20

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/2675a4020023
https://hg.mozilla.org/integration/autoland/rev/b92005ef6b2f
suppress NS_BINDING_ABORTED errors for SSE. r=necko-reviewers,valentin

Comment 21

[agoloman]

https://hg.mozilla.org/mozilla-central/rev/b92005ef6b2f