656041 - Some parts of page scroll at different rates in Fennec

Status: VERIFIED FIXED

(Core :: Graphics, defect)

Description

[Matt Brubeck (:mbrubeck)]

1. Open http://tbpl.mozilla.org/?tree=Firefox in Fennec and wait for it to finish loading.
2. Scroll the page.

Results: In my Android opt build, some parts of the content scroll normally, while others stay fixed at first and then "catch up" with the others after a delay.  (On another page where I saw this, the text scrolled normally but backgrounds lagged behind.)

In my desktop Linux debug build, I get a content process crash:

###!!! ABORT: nsDisplayScrollLayer should always be defined: 'hasCount', file /home/mbrubeck/src/mozilla/mozilla-central/layout/base/nsDisplayList.cpp, line 1938

Comment 1

[Benjamin Stover (:stechz)]

Thanks to Roc for helping me find this problem. When sets are finally merged into a stacking context, there is a particular expected ordering. First come borders and backgrounds, then negative z-index, etc. See:

http://mxr.mozilla.org/mozilla-central/source/layout/generic/nsFrame.cpp#1542

Since I was putting the info display item in Content(), it was possible for scroll layers to be put below info layers, which violates an important invariant!

Comment 2

[Benjamin Stover (:stechz)]

Attachment: Some parts of page scroll at different rates in Fennec

Comment 3

[Benjamin Stover (:stechz)]

Try build started here: http://tbpl.mozilla.org/?tree=Try&rev=50815fe4496e

Comment 4

[Martijn Wargers (dead)]

Is this a regression from bug 647192? In that case it should block that bug.

Comment 5

[Benjamin Stover (:stechz)]

Not a regression, but the reftest relies on the added displayport stuff from bug 647192.

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 531483 [details] [diff] [review]
Some parts of page scroll at different rates in Fennec

Review of attachment 531483 [details] [diff] [review]:
-----------------------------------------------------------------

Comment 10

[Matt Brubeck (:mbrubeck)]

Is this ready to land on mozilla-central?   Is it safe to request approval to land this in Aurora for Firefox 6?

Comment 11

[Matt Brubeck (:mbrubeck)]

Requesting tracking-firefox6.  This is a user-visible regression from Fennec 5 to Fennec 6.

Comment 12

[Benjamin Stover (:stechz)]

Sorry, this went off my radar. This needs another try run. I was having reftest issues but I think the FF5 regression fix might have taken care of the orange.

Comment 13

[Benjamin Stover (:stechz)]

Still seems to have an orange, sadly: http://tbpl.mozilla.org/?tree=Try&rev=af7f094914ab

Comment 14

[Benjamin Stover (:stechz)]

Fixed reftest for orange by disabling reftest for non-remote run.

Pushed to inbound http://hg.mozilla.org/integration/mozilla-inbound/rev/c64dcd99b27e

Comment 15

[Matt Brubeck (:mbrubeck)]

http://hg.mozilla.org/mozilla-central/rev/c64dcd99b27e

Comment 16

[Matt Brubeck (:mbrubeck)]

Comment on attachment 531483 [details] [diff] [review]
Some parts of page scroll at different rates in Fennec

Requesting approval-mozilla-aurora.  One-line change to fix a regression from Fx5 -> Fx6 that breaks some web pages in Fennec.  This patch has been in nightlies for the past week with no adverse effect on desktop or mobile.

Comment 17

[Johnathan Nightingale [:johnath]]

(In reply to comment #16)
> Comment on attachment 531483 [details] [diff] [review] [review]
> Some parts of page scroll at different rates in Fennec
> 
> Requesting approval-mozilla-aurora.  One-line change to fix a regression
> from Fx5 -> Fx6 that breaks some web pages in Fennec.  This patch has been
> in nightlies for the past week with no adverse effect on desktop or mobile.

(Aside: this is basically the perfect nomination comment)

Comment 18

[Doug Turner (:dougt)]

http://hg.mozilla.org/releases/mozilla-aurora/rev/034d91c4d1fc

Comment 19

[Andreea Pod]

I can still reproduce this issue following steps from bug 656041 which is duplicate of this, on Aurora build:
Mozilla /5.0 (Android;Linux armv7l;rv:6.0a2) Gecko/20110629 Firefox/6.0a2 Fennec/6.0a2
device: LG Optimus 2X (Android 2.2)

Comment 20

[Andreea Pod]

Bug 658315, sorry.

Comment 21

[Andreea Pod]

Still reproducible on Firefox 6 Beta 2: 
Mozilla /5.0 (Android;Linux armv7l;rv:6.0) Gecko/20110713 Firefox/6.0 Fennec/6.0 
following the steps from bug 658315. Reopening.

Comment 22

[Benjamin Stover (:stechz)]

Part of the problem of bug 658315 was fixed by this bug, but looking at the screenshot for bug 658315 shows that 658315 is not a duplicate of this bug. This is my error; I just saw the crash stack and assumed that it was the same problem.

Closing this bug and re-opening the dupe.

Comment 23

[Andreea Pod]

Verified fixed on Firefox 6 Beta 2: Mozilla /5.0 (Android;Linux armv7l;rv:6.0) Gecko/20110713 Firefox/6.0 Fennec/6.0

Device: LG Optimus 2X (Android 2.2)