811873 - WebRTC crashtests for mozRTCPeerConnection fail when 'browser.tabs.remote' set to true

Status: RESOLVED FIXED

(Core :: WebRTC, defect, P1)

Description

[Henrik Skupin [:whimboo][⌚️UTC+1]]

The following crashtest fails for Linux opt builds when they run the Cipc (Crashtest IPC) tests:

REFTEST TEST-UNEXPECTED-FAIL | file:///home/cltbld/talos-slave/test/build/reftest/tests/dom/media/tests/crashtests/791330.html | load failed: timed out waiting for reftest-wait to be removed

The only reason why this can happen is when the onSuccess callback is not excuted for the getUserMedia call.

Ted, do you know how we can execute those tests? Or at least who is responsible for this testrun?

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

We are also failing in '801227.html' now:

reftest/tests/dom/media/tests/crashtests/801227.html | load failed: timed out waiting for reftest-wait to be removed

I will figure out how to run those tests on Thursday, so we can get it investigated.

Comment 2

[(not currently active) Ted Mielczarek]

Sorry, meant to comment on this earlier. If you look at the full log of any test run you can always find the command used to execute the test suite. For Cipc it's this:
python reftest/runreftest.py --appname=firefox/firefox-bin --utility-path=bin --extra-profile-file=bin/plugins --symbols-path=http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-linux/1353465869/firefox-20.0a1.en-US.linux-i686.crashreporter-symbols.zip --setpref=browser.tabs.remote=true reftest/tests/testing/crashtest/crashtests.list

So the only difference from normal crashtest is the "--setpref=browser.tabs.remote=true" line.

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

That's what Randell tried but wasn't successful in it. I will check that now.

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Ok, I can reproduce this problem also on OS X now with a profile created by the reftest framework. 'browser.tabs.remote=true' doesn't seem to be the single cause here. I will analyze the profile now what specific setting or extension is causing that.

Comment 5

[Henrik Skupin [:whimboo][⌚️UTC+1]]

An exception gets thrown:

[Exception... "Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [IPeerConnection.initialize]"  nsresult: "0x80004005 (NS_ERROR_FAILURE)"  location: "JS frame :: jar:file:///Applications/Firefox/Nightly.app/Contents/MacOS/omni.ja!/components/PeerConnection.js :: PeerConnection.prototype._queueOrRun :: line 262"  data: no]

http://mxr.mozilla.org/mozilla-central/source/dom/media/PeerConnection.js#262

261     if (!this._pending) {
262       obj.func.apply(this, obj.args);
263       if (obj.wait) {

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+1]]

All tests which make use of the mozRTCPeerConnection will be affected. So this is indeed a core issue and not a test failure. We have to wait for a fix on bug 814527 now.

Comment 7

[Henrik Skupin [:whimboo][⌚️UTC+1]]

We can disable Cipc tests by just adding the following line to the crashtest list:

+skip-if(browserIsRemote) include ../../dom/media/tests/crashtests/crashtests.list

I propose that we disable all WebRTC tests for remote browsers right now. I don't think it's worth setting the flag on a case by case basis. Once we are good with those tests and multiple processes we can re-enable them.

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Attachment: Disable crashtest-ipc

Comment 9

[Randell Jesup [:jesup] (needinfo me)]

Comment on attachment 686274 [details] [diff] [review]
Disable crashtest-ipc

Review of attachment 686274 [details] [diff] [review]:
-----------------------------------------------------------------

r+; I assume you intend it to still be disabled for all tests, not just Cipc

Comment 10

[Henrik Skupin [:whimboo][⌚️UTC+1]]

(In reply to Randell Jesup [:jesup] from comment #9)
> r+; I assume you intend it to still be disabled for all tests, not just Cipc

The line in the crashtest.list file is still commented out, which is the all tests case. Keep in mind that the next merge of m-c to alder will cause a conflict here.

Comment 11

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Comment on attachment 686274 [details] [diff] [review]
Disable crashtest-ipc

https://hg.mozilla.org/integration/mozilla-inbound/rev/ea84c06a3b38

Comment 12

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/ea84c06a3b38