942830 - Scrolling randomly stops working / repainting correctly on certain pages

Status: RESOLVED FIXED

(Firefox for Metro Graveyard :: General, defect)

Description

[Dmitrij D. Czarkoff]

User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:28.0) Gecko/20100101 Firefox/28.0 (Beta/Release)
Build ID: 20131124104839

Steps to reproduce:

This problems appears randomly, most often seen on this bugzilla instance, but also randomly on itworld.com and elsewhere. With the same version of Nightly I can't reliably reproduce it, neither can get the clue about any conditions I need to secure in order to reproduce this bug.


Actual results:

Scrollability of HTML elements gets reversed: content that shouldn't move on scrolling events gets scrolled (with black bar appearing on top or on bottom of the page (depending on scrolling direction), while ordinary content (that should move on scrolling) becomes static. All means of scrolling I have (mouse wheel, scrollbar, hazardous wheel emulation with trackpoint) have the same exact effect. Previously dragging the metro firefox "window" would fix the problem for all current tabs, but with 2013/11/24 Nightly it doesn't help.


Expected results:

Everything being right.

Comment 1

[Dmitrij D. Czarkoff]

I suck at describing the UI interaction, so I would be happy to have someone fill a new bug with readable description of the problem, and marking this one as duplicate. Also: I'm updating now, so more news to come soon.

Comment 2

[Jim Mathies [:jimm]]

Maybe a dupe of bug 941972? There's a test case there you can take a look at.

Comment 3

[Jim Mathies [:jimm]]

Hmm, maybe not. on itworld.com, I'm seeing the old apzc root id changing problem.

Comment 4

[Jim Mathies [:jimm]]

Seems to show up as soon as the comment section on the right starts to load.

Comment 5

[Dmitrij D. Czarkoff]

(In reply to Jim Mathies [:jimm] from comment #2)
> Maybe a dupe of bug 941972? There's a test case there you can take a look at.

Nope, the example sites from that bugs work fine for me, and I have no means of touch input. My testcase is probably the page from ITWorld[0] where I have this bug after a couple of scrolling events every time I load the page. Survives restarts.

[0] http://www.itworld.com/operating-systems/384051/unix-debugging-your-scripts-even-more-effectively

Comment 6

[Jim Mathies [:jimm]]

Attachment: fix

Comment 7

[Jim Mathies [:jimm]]

Comment on attachment 8337798 [details] [diff] [review]
fix

This fixes the problem, but I'm concerned about cases where the page wants to consume scroll  / doesn't want built-in scroll behavior. This trap will catch scroll no matter what since we get preference in capture listeners.

What I'm not sure of yet is why we don't get the bubbled scroll event here. I assume that if itworld is calling preventDefault() on the scroll message, we wouldn't scroll by default in desktop either. That's not the case though, itworld works fine on desktop.

Comment 8

[Jim Mathies [:jimm]]

Actually, I guess they would have to call stopPropagate() to catch this vs. preventDefault(). That actually makes more sense, even if they do that, we probably keep default scroll behavior going.

If that's the case, I'm not sure how we fix this. We would want to catch the scroll event at the end of the event chain, regardless of calls to preventDefault() and stopPropagation() so we could check the preventDefault flag and act accordingly.

Comment 9

[Jim Mathies [:jimm]]

Attachment: testcase

Comment 10

[Jim Mathies [:jimm]]

I think we have some sort of uber addEventListener for chrome. I'll see if I can track that down.

Comment 11

[Jim Mathies [:jimm]]

Comment on attachment 8337798 [details] [diff] [review]
fix

smaug pointed me to - http://mxr.mozilla.org/mozilla-central/source/content/events/public/nsIEventListenerService.idl#80

Comment 12

[Jim Mathies [:jimm]]

Comment on attachment 8337798 [details] [diff] [review]
fix

I guess this is actually ok since the 'scroll' event can't be cancelled. If we get this, the page is going to be scrolled by dom, so we need to update the apzc.

Comment 13

[Jim Mathies [:jimm]]

Attachment: export patch

Comment 14

[Jim Mathies [:jimm]]

https://hg.mozilla.org/integration/fx-team/rev/2542e38220a3

Comment 15

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/2542e38220a3