629587 - page scroll synchronization with the turn of a mouse wheel is lost when mouse pointer hover over buttons. And scroll width decreases

Status: RESOLVED FIXED

(Core :: XUL, defect)

Description

[Alice0775 White]

Build Identifier:
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b11pre) Gecko/20110127 Firefox/4.0b11pre ID:20110127030333

Synchronization with the turn of a mouse wheel is lost when mouse pointer hover over buttons. And scroll width decreases.

This is the factor that scroll with the mouse wheel becomes slow.

Reproducible: Always

Steps to Reproduce:
1. Start Minefield with new profile and install much add-ons
2. Enable Smooth Scroll
3. Open Addons Manager (Ctrl+Shift+A) and Select Extensions pane

4. Put a mouse pointer on a button
5. Scroll by mouse wheel one tick

Try to compare it with the following
6. put a mouse pointer on a addon's icon
7. Scroll by mouse wheel one tick

Actual Results:
 Synchronization with the turn of a mouse wheel is lost when mouse pointer hover over buttons. And scroll width decreases

Expected Results:
 It should be scrolled in the same speed

Comment 1

[Timothy Nikkel (:tnikkel)]

Seems about the same for me on my Linux machine.

Comment 2

[Alice0775 White]

(In reply to comment #1)
> Seems about the same for me on my Linux machine.

Yes, It does not happen on Linux build.

It happens on Windows build only.

Comment 3

[Timothy Nikkel (:tnikkel)]

I wonder if this is a regression or not.

Comment 4

[Alice0775 White]

WORKAROUND: append the following CSS to userContent.css

.addon-control {
  box-shadow: none !important;
}

Comment 5

[Alice0775 White]

Attachment: sample xul

Try wheel scroll
you should enable smooth scoll

Left side box : wheel scroll is lagged
Right side box: wheel scroll is normal

Comment 6

[Alice0775 White]

The problem in comment #5 also happens on Linux.

Comment 7

[Timothy Nikkel (:tnikkel)]

Wow, thats a great test case Alice. I can see the difference on Linux. Has that testcase always acted like that or did it regress sometime I wonder.

Comment 8

[Alice0775 White]

Regression window for attachment 509754 [details](cached m-c hourly):
Works:
http://hg.mozilla.org/mozilla-central/rev/aefb8c332502
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b6pre) Gecko/20100910 Firefox/4.0b6pre ID:20100911084458
Fails:
http://hg.mozilla.org/mozilla-central/rev/26bfc0860822
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:2.0b6pre) Gecko/20100910 Firefox/4.0b6pre ID:20100911102300
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=aefb8c332502&tochange=26bfc0860822
Candidate:
Bug 590039 - fix blur radius computation and rename -moz-box-shadow to box-shadow

Comment 9

[Timothy Nikkel (:tnikkel)]

That's because "box-shadow" will have no effect prior to that change set because it wasn't recognized; only the the prefixed version -moz-box-shadow was recognized.

Comment 10

[Alice0775 White]

Attachment: sample xul -moz-box-shadow

Regression window(cached m-c hourly):
Works:
http://hg.mozilla.org/mozilla-central/rev/b94cf8147ccd
Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.3a1pre) Gecko/20100111 Minefield/3.7a1pre ID:20100111140136
Fails:
http://hg.mozilla.org/mozilla-central/rev/47cd92d616d7
Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.3a1pre) Gecko/20100111 Minefield/3.7a1pre ID:20100111150802
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b94cf8147ccd&tochange=47cd92d616d7
Candidate;
Bug 526394 - Move scrolling out of the view system into layout

Comment 11

[Timothy Nikkel (:tnikkel)]

Interesting, thank you very much Alice.

Comment 12

[Timothy Nikkel (:tnikkel)]

This might be the explanation for the mysterious scrolling bugs (slow scrolling, laggy scrolling, jerky scrolling, amount of scrolling decreased) we have been seeing.

Comment 13

[Timothy Nikkel (:tnikkel)]

The first bad revision is:
changeset:   37083:4ccff5df452c
user:        Robert O'Callahan <[e-mail]>
date:        Thu Oct 08 16:01:15 2009 +1300
summary:     Bug 526394. Part 31: Move scroll implementation into nsGfxScrollFrame. r=mats

Comment 14

[Timothy Nikkel (:tnikkel)]

smooth scroll just makes this more obvious, it happens without it.

Comment 15

[Timothy Nikkel (:tnikkel)]

Attachment: patch

We reflow while smooth scrolling and nsGfxScrollFrameInner::CurPosAttributeChanged gets called from the scroll frames ReflowFinished reflow callback. It gets the position information from the scrollbar and then calls ScrollTo(dest, INSTANT) because the scrollbar does not have the smooth attribute (it didn't on 3.6 either). That cancels the inprogress async smooth scroll. dest is actually the current position of the scroll frame anyway, so the scroll call is useless, but it only determines that after cancelling the async scroll. http://hg.mozilla.org/mozilla-central/rev/4ccff5df452c removed a "if (x == curPosX && y == curPosY) return;" check in nsGfxScrollFrameInner::CurPosAttributeChanged. I think we should just restore it.

This patch also improves the behaviour when I scroll down quickly with the mouse wheel and smooth scrolling is turned off.

Comment 16

[Timothy Nikkel (:tnikkel)]

The reason that non-smooth scrolling is improved is that we still use an async scroll with a timer to do the scrolling, so if we get a couple of nsGfxScrollFrameInner::ScrollTo calls without the async timer happening and then a reflow (off the refresh driver or will paint) it will end up in nsGfxScrollFrameInner::CurPosAttributeChanged and try to scroll INSTANT to the current position even though there is an async scroll object pending for many pixels further down.

The following dump makes it plain.

nsGfxScrollFrameInner::ReflowFinished calling CurPosAttributeChanged
I'm nsGfxScrollFrameInner::CurPosAttributeChanged and i'm calling ScrollTo
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 0 (0 0) mode INSTANT
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 51 (0 3060) mode SMOOTH
nsGfxScrollFrameInner::AsyncScrollCallback calling ScrollToImpl with dest 0 51 (0 3060)
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 102 (0 6120) mode SMOOTH
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 153 (0 9180) mode SMOOTH
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 204 (0 12240) mode SMOOTH
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 255 (0 15300) mode SMOOTH
nsGfxScrollFrameInner::ReflowFinished calling CurPosAttributeChanged
I'm nsGfxScrollFrameInner::CurPosAttributeChanged and i'm calling ScrollTo
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 51 (0 3060) mode INSTANT
nsGfxScrollFrameInner::ScrollTo aScrollPosition 0 102 (0 6120) mode SMOOTH
nsGfxScrollFrameInner::AsyncScrollCallback calling ScrollToImpl with dest 0 102 (0 6120)

Comment 17

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

Right. OK. Great!

Comment 18

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/mozilla-central/rev/2ba4709dc2ce

Comment 19

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

I don't think anything still needs landing here.