Firefox hang when clicking an item in https://store.docker.com/search

NEW
Unassigned

Status

()

Core
Disability Access APIs
P2
critical
7 months ago
2 months ago

People

(Reporter: kodmivi, Unassigned)

Tracking

(Depends on: 1 bug, Blocks: 1 bug, {hang})

53 Branch
x86_64
Windows 7
Points:
---
Dependency tree / graph

Firefox Tracking Flags

(firefox-esr45 unaffected, firefox-esr52 fix-optional, firefox53 wontfix, firefox54 affected, firefox55 affected)

Details

(Reporter)

Description

7 months ago
This bug was filed from MDN. Firefox is hanging, and manually crashing using crashfirefox64.exe produces this crash report: https://crash-stats.mozilla.com/report/index/622b6e18-cc02-4686-b722-9d53d1170427

Steps to reproduce
1. Open https://store.docker.com/search?operating_system=windows&q=&source=verified&type=image
2. Click on 'nanoserver' item.
(Reporter)

Comment 1

7 months ago
More details:
- When hanging, Firefox consumes 100% of CPU core.
- The stack-trace of the hanging thread contains 'mozilla::a11y::NotificationController::CoalesceMutationEvents()' function call.
Summary: Firefox hang → Firefox hang when clicking an item in https://store.docker.com/search
Based on the stack mentioned in comment #1, this looks like an a11y bug to me.
Component: General → Disability Access APIs
Product: Firefox → Core

Comment 3

7 months ago
Alex, have you seen this before?
Flags: needinfo?(surkov.alexander)

Comment 4

7 months ago
Nope. CoalesceMutationEvents is a known bottle neck though. I'm curious if Firefox 48, which has bug 1261425, is any better.

In general I think we should try to go towards to pruning events in subtree (like bug 1301829).
Blocks: 531850
Flags: needinfo?(surkov.alexander)
(In reply to alexander :surkov from comment #4)
> Nope. CoalesceMutationEvents is a known bottle neck though. I'm curious if
> Firefox 48, which has bug 1261425, is any better.
> 
> In general I think we should try to go towards to pruning events in subtree
> (like bug 1301829).

The crash/hang is from 53.

Comment 6

7 months ago
I attempt finding regression window with Windows10 64bit Nightly 32bit with e10s disabled + a11y activated w/Japanese IME ATOK2016.

#1 Regression window(Hang at step 1):
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=850159471c686be76390a2ee8dae12afbac7cc32&tochange=f38aaa49ed11b2e20bced1334e44e203b29e1040

Regressed by: Bug 1296420

#2 Regression window(Hang at Step 2)
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=ef23bf8110953349a23ba2e344c20643b392eda0&tochange=387d3acae9e99bdc140a65fd367ecbaa6238f3a3

Regressed (partially progressed?) by: Bug 1270916


@Michael Li and @:tbsaunde,
These patches cause the hang. Please look this?
Blocks: 1296420, 1270916
No longer blocks: 531850
Status: UNCONFIRMED → NEW
status-firefox53: --- → wontfix
status-firefox54: --- → affected
status-firefox55: --- → affected
status-firefox-esr45: --- → unaffected
status-firefox-esr52: --- → fix-optional
Ever confirmed: true
Flags: needinfo?(tbsaunde+mozbugs)
Flags: needinfo?(michael.li11702)
Keywords: hang

Updated

7 months ago
Blocks: 531850

Comment 7

7 months ago
(In reply to :Gijs from comment #5)
> (In reply to alexander :surkov from comment #4)
> > Nope. CoalesceMutationEvents is a known bottle neck though. I'm curious if
> > Firefox 48, which has bug 1261425, is any better.
> > 
> > In general I think we should try to go towards to pruning events in subtree
> > (like bug 1301829).
> 
> The crash/hang is from 53.

I can reproduce the hang on 52esr w/ e10s disabled + a11y activated. The combination is default for ATOK2016 IME user.

Comment 8

7 months ago
Michael Li was an intern and is currently no longer with us, and Tbsaunde is on PTO. Alex, see comment #6 for the regression windows.
Flags: needinfo?(tbsaunde+mozbugs)
Flags: needinfo?(surkov.alexander)
Flags: needinfo?(michael.li11702)

Comment 9

7 months ago
(In reply to Alice0775 White from comment #6)
> I attempt finding regression window with Windows10 64bit Nightly 32bit with
> e10s disabled + a11y activated w/Japanese IME ATOK2016.
> 
> #1 Regression window(Hang at step 1):
> https://hg.mozilla.org/integration/mozilla-inbound/
> pushloghtml?fromchange=850159471c686be76390a2ee8dae12afbac7cc32&tochange=f38a
> aa49ed11b2e20bced1334e44e203b29e1040
> 
> Regressed by: Bug 1296420

this bug is surprising with me, cannot think how it may make the event coalescence worse.

> #2 Regression window(Hang at Step 2)
> https://hg.mozilla.org/integration/mozilla-inbound/
> pushloghtml?fromchange=ef23bf8110953349a23ba2e344c20643b392eda0&tochange=387d
> 3acae9e99bdc140a65fd367ecbaa6238f3a3
> 
> Regressed (partially progressed?) by: Bug 1270916

ok, that's what I concerned about in 1270916 comment #c73, when switched from EventTree coalescence (introduced in bug 1261425).

I'm also curious if single process Firefox is same bad as e10s one.
Flags: needinfo?(surkov.alexander)

Updated

6 months ago
Depends on: 1368784

Updated

4 months ago
Priority: -- → P1
Priority: P1 → P2
You need to log in before you can comment on or make changes to this bug.