Closed Bug 1546850 Opened 5 years ago Closed 5 years ago

Intermittent leakcheck | tab 37416 bytes leaked (APZEventState, ActiveElementManager, AtomSet, BackstagePass, BrowserChild, ...)

Categories

(Remote Protocol :: CDP, defect, P2)

defect

Tracking

(Not tracked)

RESOLVED DUPLICATE of bug 1546945

People

(Reporter: intermittent-bug-filer, Unassigned)

References

(Regression)

Details

(Keywords: intermittent-failure, regression, Whiteboard: [stockwell unknown])

#[markdown(off)]
Filed by: nbeleuzu [at] mozilla.com

https://treeherder.mozilla.org/logviewer.html#?job_id=242466222&repo=autoland

https://queue.taskcluster.net/v1/task/HCPX-EHVTX2NJE1_KPOdBg/runs/0/artifacts/public/logs/live_backing.log

After the backout of Bug 1527287 , the leak changed from BrowsingContext to BrowserChild.
:eakhgari , can you please take a look?

Flags: needinfo?(ehsan)
Priority: -- → P2

Ryan Hunt renamed this class in bug 1534395. If you want to blame someone for the object name changing in the leak report (just joking of course) it'd be him. ;-)

But of course this is the dupe of whatever bug was tracking the leak with the old name of the object (perhaps bug 1527087 but I am not sure.) Unccing myself because that is all I know here.

Flags: needinfo?(ehsan)

Ryan can you please take a look at this?

There are 146 failures in the last 7 days on linux32, linux64, osx-10-10, windows10-64 and windows7-32 all debug builds.

Recent failure log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=242716950&repo=mozilla-inbound&lineNumber=7704

Flags: needinfo?(rhunt)
Regressed by: 1534395

I looked at the recent failure log from comment 4, and it is happening in the directory remote/test/browser/, so I think the recent spike in failures is more likely to be the leakcheck variant of bug 1547303, which also started on 4/24.

Component: DOM: Core & HTML → Target
Flags: needinfo?(rhunt) → needinfo?(ato)
Product: Core → Remote Protocol
Regressed by: 1546385
No longer regressed by: 1534395

I looked at about 1 out of 10 of the failure logs from comment 4, and I'd say 90% or more of them are in remote/test/browser/. There were also a couple of failures in service worker related tests, but it didn't seem that bad.

We disabled remote/test/browser/browser_cdp.js because it perma-failed
the leakcheck in bug 1546945 and we thought it was limited to that
test specifically. According to this and bug 1547303, it appears
to happen also for the other tests in the same directory, though
less frequently.

We should disable all the browser-chrome tests under remote/test/browser/
until we’ve had time to plug the leak.

Status: NEW → RESOLVED
Closed: 5 years ago
Flags: needinfo?(ato)
Resolution: --- → DUPLICATE
Component: CDP: Target → CDP
Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.