818345 - Crash: Swipe to close for tabs causes ArrayIndexOutOfBounds on Froyo and Gingerbread

Status: RESOLVED WORKSFORME

(Firefox for Android Graveyard :: Theme and Visual Design, defect)

Description

[Chenxia Liu [:liuche]]

Attachment: Log: adb output of crash.

On Nightly 20.0a1 (2012-12-04)

Swiping to close tabs causes a crash, on an Android 2.2.2 device.

STR:
1. Open Nightly with multiple tabs.
2. View all tabs by clicking # in top right.
3. Swipe to close one of the tabs.

Expected:
Tab would be closed, can continue using Nightly.

Actual:
Nightly crashes, launches crash reporter.

Comment 1

[Chenxia Liu [:liuche]]

Crash report: https://crash-stats.mozilla.com/report/index/8b3fd8d7-a5e5-41a9-8e0f-2bc1d2121205

Comment 2

[Aaron Train [:aaronmt]]

2.2.x only? I'm unable to reproduce on 4.1.2/4.0.4. Haven't tried 2.3.

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I started seeing this constantly just recently (maybe because the [X] to close tabs was removed and swipe-to-close is the only mechanism?).

In any case, I definitely recall seeing a bug like this before (and which was fixed) relation to MotionEvent.getY, exceptions, and old Android versions, but I can't find it.

Comment 4

[Kevin Brosnan [Ex-Mozilla]]

Can any of the SF devs take a look at these phones?

Comment 5

[Mark Finkle (:mfinkle) (use needinfo?)]

Wes or Sriram would be good candidates for taking a look.

Comment 6

[Chris Peterson [:cpeterson]]

I can take a look at David's phone. I have another Froyo phone, too.

The bug looks like a race condition because swipe's touch event causes an ArrayIndexOutOfBoundsException, suggesting the touch target (in the tab ListView) was removed after the touch event was queued but before it was processed.

Comment 7

[Chris Peterson [:cpeterson]]

David's phone does not give any more info than the stack traces we already have. As I noted in comment 6, I think this is a race condition related to the order in which we process swipe MotionEvents and unload the tabs ListView.

Comment 8

[Lukas Blakk [:lsblakk] use ?needinfo]

Since this is reproducible and we've got an idea of where to start looking, tracking and assigning to Mark for delegation.

Comment 9

[Mark Finkle (:mfinkle) (use needinfo?)]

Pushing to Wes

Comment 10

[Wesley Johnston (:wesj)]

just had david show me his phone and this is no longer happening there. I also don't see any reports after 20.0b2 (and its very few of the crashes there). I wonder if some of our threading fixes have helped (but i see bug 844407 isn't fixed yet).

Since 19 is shipping and this is fixed in 20 it seems, is it worth chasing down?

Comment 11

[Wesley Johnston (:wesj)]

Grr. My crash-stats foo is not great. There are reports of this on 21 and one on 22. Digging a little more...

Comment 12

[Scoobidiver (away)]

More reports at https://crash-stats.mozilla.com/report/list?signature=java.lang.ArrayIndexOutOfBoundsException%3A+at+android.view.MotionEvent.getY%28MotionEvent.java%29

Comment 13

[Robert Kaiser]

(In reply to Wesley Johnston (:wesj) from comment #11)
> Grr. My crash-stats foo is not great. There are reports of this on 21 and
> one on 22. Digging a little more...

Those seem to have stopped as well, and nothing new on 20 post-b2. Sounds to me like we can untrack for that release.

Comment 14

[bhavana bajaj [:bajaj]]

(In reply to Robert Kaiser (:[e-mail]) from comment #13)
> (In reply to Wesley Johnston (:wesj) from comment #11)
> > Grr. My crash-stats foo is not great. There are reports of this on 21 and
> > one on 22. Digging a little more...
> 
> Those seem to have stopped as well, and nothing new on 20 post-b2. Sounds to
> me like we can untrack for that release.

Untracking for 21 based on this.