715856 - [10.7] Trying to leave Mibbit site with 2 finger swipe succeeds prematurely, and causes a modal dialog "hang"

Status: RESOLVED WORKSFORME

(Core :: General, defect)

Description

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7; rv:11.0a2) Gecko/20111229 Firefox/11.0a2 ID:20111229042021

If the user has configured Firefox to use the 2 finger swipe to navigate back and forward, the browser will completely hang due to the modal alert. Steps to reproduce are really simple:

Steps:
1. Open https://wiki.mozilla.org/QA/Automation_Services/Team
2. Click on one of the IRC links and use Mibbit to connect to IRC
3. Wait until Mibbit has been loaded and swipe with 2 fingers to the left
4. Try to close the modal alert

We will not be successful with step 4. Actually the browser has been frozen and needs to be force killed.

I can reproduce it back to Firefox 9 for now. Earlier versions somehow don't react on the two finger swipe and don't show this behavior.

Could that be related to the new implementation which initially went into Firefox 9 by bug 668953?

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Not sure but it could be that I hit bug 715829 due to this bad behavior.

Comment 2

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Would be great if someone could test Firefox 9 on 10.6 if there is the same issue or not.

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

This issue doesn't occur when you close the tab, or click onto the back button. So it seems like we have an issue with the event loop?

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Also this only happens for me when 'Swipe between full-screen apps' is disabled in the trackpad / mouse system settings.

Comment 5

[Henrik Skupin [:whimboo][⌚️UTC+1]]

As talked with Virgil on IRC this is not reproducible in 10.6 and that's why related to the two finger swipe implementation in 10.7.

Comment 6

[Virgil Dicu [:virgil] [QA]]

Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7; rv:11.0a2) Gecko/20120105 Firefox/11.0a2

Used whimboo's steps from comment 0.

CAN reproduce with any of the following settings for "swipe between pages":
-"swipe left or right with one finger"
-"swipe with one or two fingers"

CAN'T reproduce with the following setting for "swipe between pages":
-"swipe left or right with two fingers"

Comment 7

[Steven Michaud [:smichaud] (Retired)]

I can reproduce this with default System Preferences settings
(including with "Swipe between pages" set to "Scroll left or right
with two fingers").

The "hang" is bug 476541, which for complicated reasons isn't going to
get fixed anytime soon.  For more information see bug 691702 comment
#8 and bug 691702 comment #9.

But the actual bug is that two-finger swipes, as currently
implemented, can take place even when "blocked" by a modal dialog.
This shouldn't be too hard to fix.  That's what	I'll be	doing.

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+1]]

(In reply to Steven Michaud from comment #7)
> The "hang" is bug 476541, which for complicated reasons isn't going to
> get fixed anytime soon.  For more information see bug 691702 comment
> #8 and bug 691702 comment #9.

Huh, why is that bug 476541? There is no other modal dialog open. It's really just the dialog from Mibbit.

Also this is a regression not from the original implementation of the 2 finger swipe. Working on the regression range right now. Currently it happened between 111001 and 111007.

> But the actual bug is that two-finger swipes, as currently
> implemented, can take place even when "blocked" by a modal dialog.
> This shouldn't be too hard to fix.  That's what	I'll be	doing.

I still think the hang is different and needs to be fixed too.

Comment 9

[Steven Michaud [:smichaud] (Retired)]

> Also this is a regression not from the original implementation of
> the 2 finger swipe. Working on the regression range right
> now. Currently it happened between 111001 and 111007.

I'll await your results, and see if I can reproduce them.

Comment 10

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Regressed for me between the builds 2011100303 and 2011100403.

WORKS: http://hg.mozilla.org/mozilla-central/rev/90575e23ea93
FAILS: http://hg.mozilla.org/mozilla-central/rev/149fc4a6efca

Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=90575e23ea93&tochange=149fc4a6efca

I could imagine the following two bugs have an influence: bug 623625, bug 686469

Comment 11

[Steven Michaud [:smichaud] (Retired)]

I can reproduce your regression range, Henrik.

But my results with the 2011-10-04 mozilla-central nightly are different from my results with today's mozilla-central nightly.  Both "hang" (unlike the 2011-10-03 nightly).  But today's nightly also swipes back to the previous page even while the modal dialog is still up that asks if the user wants to leave the current page.

This behavior starts with the 2011-10-06 mozilla-central nightly (and isn't present in the 2011-10-05 nightly).

It's probably unrelated, but I'm going to find out which patch triggered this.

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Ok, so the hang is visible since the back-out for bug 622315 happened. Looks like that with the implementation of the two-finger swipe and the patch for bug 622315 we don't see this hang, but it would probably manifest itself once we back-out it locally for earlier builds until the changeset for the two-finger swipe implementation on bug 668953.

So not sure I can do more on the regression test front. More helpful would probably be a minimized testcase here.

Comment 13

[Justin Lebar (not reading bugmail)]

The patch for bug 622315 basically said, "You may not navigate while a navigation is pending."  So we're probably getting confused somehow here.

Comment 14

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Attachment: testcase (zip archive)

Minimized testcase from Mibbit. Sadly I cannot strip down the remaining JS code because stuff is getting dynamically created.

Comment 15

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Steven, do you have any update regarding the tests you wanted to do?

Comment 16

[Steven Michaud [:smichaud] (Retired)]

Right now I'm working on bug 715885, which is a topcrasher and seems more urgent.  I should be able to get back to this bug later this week.

Comment 17

[Alex Keybl [:akeybl]]

[Triage Comment]
Not enough volume of reports to warrant tracking. We would, however, consider uplifting if a low-risk patch is landed on m-c.

Comment 18

[Steven Michaud [:smichaud] (Retired)]

Henrik, can you still reproduce this?

Comment 19

[Henrik Skupin [:whimboo][⌚️UTC+1]]

No, I can't with the latest version of Aurora. Some landed patch might have fixed that. Closing as WFM.