Closed Bug 1574567 Opened 1 year ago Closed 1 year ago

Frequent "Assertion failure: parentFound" with Fission enabled

Categories

(Core :: DOM: Service Workers, task, P2)

task

Tracking

()

RESOLVED WORKSFORME
Fission Milestone M4.1

People

(Reporter: mccr8, Unassigned)

References

(Blocks 1 open bug)

Details

Attachments

(1 obsolete file)

There are 11 tests that are annotated that they crash with Fission with some service worker manager service related crash, but they don't actually crash, so we should enable them. I think there was one other service worker related test that I noticed passes with Fission.

browser_antitracking_subiframes.js is the test that passes now.

browser_antitracking_subiframes.js also seems to pass.

And browser_permissions_event_telemetry.js.

Pushed by amccreight@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/2bd128c84b4e
Enable various tests for Fission that don't actually crash with ServiceWorkerManagerService signatures. r=kmag
Status: NEW → RESOLVED
Closed: 1 year ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla70

Somehow I failed to realize my try run only included Mochitest plain, which explains how I didn't notice this on my try run...

This is pretty bizarre, though. None of the test directories that my patch touches seem to be running in bc2.

The crash seems to happen in shutdown after tests in browser/base/content/test/trackingUI/ run. Maybe bc tests aren't run-by-directory and my patch caused bucketing to shift somehow? Is this kind of weirdness familiar to you, Kris?

Flags: needinfo?(continuation) → needinfo?(kmaglione+bmo)
Regressions: 1574939
No longer regressions: 1574939

Ok, so there's just the intermittent leak from comment 6 to address. That looks service worker related, so it makes more sense. I'll see if I can reproduce that today.

Flags: needinfo?(kmaglione+bmo)
Fission Milestone: --- → M4

On second look, it seems like there are a number of crash that match what Kris filed.

They are actually assertion failures:
Assertion failure: parentFound, at /builds/worker/workspace/build/src/dom/serviceworkers/ServiceWorkerManagerService.cpp:188

Please back this patch out.

Status: RESOLVED → REOPENED
Keywords: checkin-needed
Resolution: FIXED → ---
Target Milestone: mozilla70 → ---

Backout by csabou@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/a6a6b91c035e
Backed out changeset 2bd128c84b4e for causing assertion failures on ServiceWorkerManagerService.cpp.

Keywords: checkin-needed

I looked through the failures in comment 6, and they break down like this:

  • Assertion failure: parentFound, at /builds/worker/workspace/build/src/dom/serviceworkers/ServiceWorkerManagerService.cpp:188
    4 browser/base/content/test/permissions/browser_permissions.js
    1 browser/base/content/test/permissions/browser_temporary_permissions_expiry.js
    1 browser/base/content/test/permissions/browser_temporary_permissions_navigation.js

  • browser/base/content/test/permissions/browser_autoplay_blocked.js | Test timed out

  • leakcheck | tab 488 bytes leaked (ServiceWorkerJob, ServiceWorkerJobQueue, ServiceWorkerJobQueue::Callback, nsAuthURLParser, [...]
    browser/base/content/test/sanitize/browser_sanitize-offlineData.js

  • leakcheck | tab 7852 bytes leaked (ChannelEventQueue, CondVar, ConsoleReportCollector, CookieSettings, HttpChannelChild, ...)
    2 browser/base/content/test/permissions/browser_temporary_permissions_expiry.js
    browser/base/content/test/permissions/browser_permissions.js

I also looked at a few other M-fis() bc runs, and saw "Assertion failure: parentFound" in the other tests I disabled.

From the browser.ini annotation, browser_devtools_serviceworker_interception.js also seems to hit this assertion.

This assertion failure also seems to happen on trunk, but much less frequently, as seen in in bug 1522243.

Depends on: 1522243
Summary: Enable various tests for Fission that don't actually crash with ServiceWorkerManagerService signatures → Frequent "Assertion failure: parentFound" with Fission enabled

Andrew, it looked like you did at least some basic analysis of this assertion failure in bug 1522243. Do you have time to look at why the failure rate rate is a lot higher in Fission, or maybe it is expected right now? Thanks.

Flags: needinfo?(bugmail)
Depends on: 1456995
Flags: needinfo?(bugmail)
Priority: -- → P2
Assignee: continuation → nobody
Attachment #9086127 - Attachment is obsolete: true

Roll some unfixed bugs from Fission Milestone M4 to M5

0ee3c76a-bc79-4eb2-8d12-05dc0b68e732

Fission Milestone: M4 → M5

M4.1 and WFM

kmag thinks this bug may have been fixed by enabling ServiceWorker e10s.

Status: REOPENED → RESOLVED
Fission Milestone: M5 → M4.1
Closed: 1 year ago1 year ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.