Page MenuHomePhabricator

WebSocket error occurs when navigating phabricator pages
Closed, WontfixPublic

Description

Firefox 36, OSX 10.9

Notifications continue to work properly, despite the error. This may not be a phabricator problem, but actually an issue with Firefox - this does not occur in Safari or Chrome.

Navigating to phabricator website (this includes my local test installs both with and without SSL, and on secure.phabricator.com website), turn on the developer console and refresh the page. An error similar to the following appears:

The connection to wss://phab.company.com:22280/ was interrupted while the page was loading.

Even though the error states "while the page was loading" I believe it's an issue with the connection not being cleaned up when the page is being navigated away from. This stackoverflow may be related:

http://stackoverflow.com/questions/14140414/websocket-interrupted-while-page-is-loading-on-firefox-for-socket-io

Related Objects

Event Timeline

cspeckmim raised the priority of this task from to Needs Triage.
cspeckmim updated the task description. (Show Details)
cspeckmim added a subscriber: cspeckmim.
chad renamed this task from WebSocket error occurs when navigating phabricator pages. to WebSocket error occurs when navigating phabricator pages.Mar 3 2015, 11:15 PM
chad added a project: Notifications.

At least at one point in time, registering anything on beforeunload would disable the "bfcache":

http://stackoverflow.com/questions/18019418/bfcache-disabled-by-unload-events-back-forward-cache

If that's still the case, we might be looking at a choice between a slow back button without a log warning, or fast back button with a log warning. Fast + log is clearly better there since this doesn't do anything bad.

Quicksand may also mitigate this somewhat by reducing the number of real page loads we perform.

epriestley triaged this task as Wishlist priority.Mar 4 2015, 2:31 PM

I removed this as a blocker for shipping Quicksand, though shipping Quicksand may help obviate this issue since we will do less page loads.

epriestley claimed this task.

I don't think this is worth fixing in general. It doesn't do anything bad (just dump a warning to a developer log which users normally don't see) and fixing it isn't trivial.