WebrtcEventLogApiTest tests failing on Site Isolation FYI bot |
||||||
Issue descriptionThe following WebRTC tests are failing on the Site Isolation FYI bot, which runs tests with the --site-per-process flag (to enable out-of-process iframes for cross-site frames): WebrtcEventLogApiTest.TestStartTimedWebRtcEventLogging WebrtcEventLogApiTest.TestStartStopWebRtcEventLogging They were introduced in https://codereview.chromium.org/1650133002 and started failing right away: https://build.chromium.org/p/chromium.fyi/builders/Site%20Isolation%20Linux/builds/8775 Looks like a JavaScript error is happening. I'll paste a full log failure below. ../../chrome/browser/media/webrtc_browsertest_base.cc:133: Failure Value of: hit_javascript_errors_.Get() Actual: true Expected: false I'll disable the tests on that bot for now, but please fix or revert as soon as you can. Thanks!
,
Apr 13 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/d5081eb2f8ad8c174b109655e28eb452231571ee commit d5081eb2f8ad8c174b109655e28eb452231571ee Author: lukasza <lukasza@chromium.org> Date: Wed Apr 13 21:04:21 2016 Adding site-per-process exceptions for WebrtcEventLogApiTest.TestStart tests. BUG= 603261 NOTRY=true TBR=creis@chromium.org, terelius@chromium.org Review URL: https://codereview.chromium.org/1885003003 Cr-Commit-Position: refs/heads/master@{#387084} [modify] https://crrev.com/d5081eb2f8ad8c174b109655e28eb452231571ee/testing/buildbot/filters/site-per-process.browser_tests.filter
,
Apr 14 2016
The javascript exception seems to come from the call
data = chrome.embeddedSearch.searchBox.getMostVisitedItemData(args.rid);
in chrome/browser/resources/local_ntp/most_visited_single.js, line 288
The call returns undefined, so accessing its members cause an exception on line 289. This part of the code is unrelated to my change.
One easy work-around would be to add an
if (data) {
// Use data do add tile as in existing code.
} else {
tiles.appendChild(renderTile(null));
}
but I'm not sure whether this is the best way. I'd also like to understand why my test trigger this problem but the similar PeerConnection test does not.
,
Apr 14 2016
What PeerConnection test are you referring to? It's possible it's disabled as well. We have a few known issues with the NTP in --site-per-process mode, such as issue 456420. It's possible that's to blame if you're loading the NTP in a test, but I'm curious about similar tests if they're not disabled for us already in site-per-process.browser_tests.filter.
,
Apr 15 2016
I was thinking about WebRtcBrowserTest.RunsAudioVideoWebRTCCallInTwoTab, and you're right; that test is actually disabled on the site isolation bot.
,
Apr 15 2016
Ah, that makes more sense. I'll mark this as blocked on issue 456420, and I'll see if we can re-enable it once that's fixed. If there are still issues with it after that, I'll let you know.
,
Apr 20 2016
,
Jun 1 2016
Moving this nonessential bug to the next milestone. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Jul 6 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/b1320e8f04eb906f295def43d4cf4894d7bc5b7a commit b1320e8f04eb906f295def43d4cf4894d7bc5b7a Author: alexmos <alexmos@chromium.org> Date: Wed Jul 06 17:38:17 2016 Clean up the browser_tests filter for Site Isolation bots. Previously, the filter disabled 192 tests. However, lots of these tests were actually passing with --site-per-process, mostly due to unnecessarily wide wildcards. This CL explicitly lists all the actual failures, and also removes a couple of WebRtc tests which seem to be passing now. This brings the number of disabled tests down to 26. BUG=417518, 603261 CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_site_isolation Review-Url: https://codereview.chromium.org/2120673002 Cr-Commit-Position: refs/heads/master@{#403928} [modify] https://crrev.com/b1320e8f04eb906f295def43d4cf4894d7bc5b7a/testing/buildbot/filters/site-per-process.browser_tests.filter
,
Jul 6 2016
I've recently discovered that these tests are passing with --site-per-process, so I've reenabled them on our FYI bots in #9. Not sure what fixed them; the NTP issue 456420 is still not fixed, but perhaps these tests stopped relying on that? Anyways, marking this as fixed for now, and let's reopen if this reoccurs.
,
Jan 24 2018
|
||||||
►
Sign in to add a comment |
||||||
Comment 1 by creis@chromium.org
, Apr 13 2016