Crash when panning/zooming leaflet map site with many markers

RESOLVED INCOMPLETE

Status

()

--
critical
RESOLVED INCOMPLETE
7 years ago
5 months ago

People

(Reporter: wlach, Unassigned)

Tracking

({crash, reproducible})

Trunk
ARM
Android
crash, reproducible
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

(Whiteboard: [native-crash], URL)

Attachments

(1 attachment, 1 obsolete attachment)

I just knocked up a mobile version of my nixi site (http://nixi.ca) using leaflet.js and tried loading it with Firefox for Android to see what would happen.

Here's the URL: http://nixi.ca/mobile.html

(I'll try to keep it up for a while)

Performance in general is janky, but more importantly it's very easy to trigger a crash just by panning/zooming in on the page. Just move to the center and pinch-zoom in and out repeatedly.

This is reproducible on all versions of Firefox for Android I tried (beta, Aurora, and Nightly).
Can you provide the crash ID from about:crashes?
Severity: normal → critical
Keywords: crash
Whiteboard: [native-crash]
1. I forgot to mention that this is on a Galaxy Nexus running Android 4.1. Haven't tried other platforms yet.

2. Weirdly enough, I don't see these crashes in about:crashes? Maybe it's due to some kind of out-of-memory condition that prevented the crash from being reported.

3. I also just realized that the example I put up wouldn't really work outside of Montreal. I put up a new version here:

http://people.mozilla.com/~wlachance/leaflet-crasher/index.html

4. I also did some digging and found that people have had similar problems with other sites and the android stock browser:

https://github.com/CloudMade/Leaflet/issues/909

There's a suggestion in the bug to set `window.L_DISABLE_3D = true` before including Leaflet (to disable hardware acceleration). Doing so does indeed seem to fix the crash here.
At least, do you have a logcat?
Keywords: reproducible
Created attachment 654228 [details]
Logcat just before crash
Attached a logcat, nothing really interesting that I could see. I also ran top to see what was going on:

 8289  1  22% S    40 734792K 203400K  fg u0_a60   org.mozilla.fennec_aurora
 8289  1   1% S    40 726152K 184512K  fg u0_a60   org.mozilla.fennec_aurora
 8289  0  39% S    40 741752K 208488K  fg u0_a60   org.mozilla.fennec_aurora
 8289  0  52% S    40 760212K 235508K  fg u0_a60   org.mozilla.fennec_aurora
 8289  0  52% S    40 720268K 184020K  fg u0_a60   org.mozilla.fennec_aurora
 8289  0  55% S    40 955752K 378488K  fg u0_a60   org.mozilla.fennec_aurora

It looks like RSS shoots right up just before the crash.
If it is caused by the screenshot code (as you surmised in your blog post) then bug 791263 should help here.
Created attachment 766373 [details] [diff] [review]
Part 1: Define state directory in mach context object

I'm gonna take a quick stab at this. But, it requires a new feature in
the mach core.
Attachment #766373 - Flags: review?(jhammel)
Assignee: nobody → gps

Comment 8

6 years ago
Comment on attachment 766373 [details] [diff] [review]
Part 1: Define state directory in mach context object

lgtm; some comments might be nice
Attachment #766373 - Flags: review?(jhammel) → review+
I accidentally. Bug 794580.
Assignee: gps → nobody
For what it's worth, this site still reliable crashes fennec, and doesn't produce any bug reports.

Updated

6 years ago
Attachment #766373 - Attachment is obsolete: true
Let's just resolve this bug as incomplete, the reproduction steps are no longer valid.
Status: NEW → RESOLVED
Last Resolved: 5 months ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.