Bug 62383 – [Chromium] Overlay scrollbars leave glitches on web content

• Component: Platform
• Status: RESOLVED FIXED

[Sailesh Agrawal]

Scrolling a page with overlay scrollbars leaves trailing artifacts on web content.

This needs to be fixed before overlay scrollbars are enabled.

[Sailesh Agrawal]

Created attachment 97201 [details]
Patch

[Sailesh Agrawal]

This doesn't fix the root cause, any suggestions appreciated.

[Sailesh Agrawal]

Created attachment 97213 [details]
Patch

[Nico Weber]

Comment on attachment 97213 [details]
Patch

don't set r? on work-in-progress patches

[James Robinson]

Comment on attachment 97213 [details]
Patch

How does the PLATFORM(MAC) port handle this in the non-composited case?

Whatever the answer is here it does not belong in ScrollView.cpp

[Sailesh Agrawal]

(In reply to comment #5)
> (From update of attachment 97213 [details])
> How does the PLATFORM(MAC) port handle this in the non-composited case?

I think that on PLATFORM(MAC) this is handled by the native host scroll view. Our RenderWidget just blindly copies the scrollrect even for overlay scrollbars.

> Whatever the answer is here it does not belong in ScrollView.cpp

What do you think of fixing this in RenderWidget. This would be a little more complicated because I'd have to pipe more information about scrollbars to RenderWidget.

[Nico Weber]

sail: What was the file where you removed the ?: earlier today? Did that help?

[Sailesh Agrawal]

(In reply to comment #7)
> sail: What was the file where you removed the ?: earlier today? Did that help?

That was in ScrollView.cpp ScrollView::rectToCopyOnScroll(). Changing it didn't help.

[James Robinson]

The mac port doesn't use platform widgets for scrolling in WebKit2

[Sailesh Agrawal]

(In reply to comment #9)
> The mac port doesn't use platform widgets for scrolling in WebKit2

Hm... how can I tell if I'm running Safar in composited or non-composited mode? Maybe they have this bug too.

[Sailesh Agrawal]

Created attachment 97369 [details]
Patch

[Sailesh Agrawal]

(In reply to comment #9)
> The mac port doesn't use platform widgets for scrolling in WebKit2

K, I think I got it. WebChromeClient::scroll() intersects the scroll rect with the clip rect. This excludes the scrollbars when overlay scrollbars are enabled.

I'm running try bots now.

[Sailesh Agrawal]

Try bot results look good.

http://build.chromium.org/p/tryserver.chromium/builders/linux/builds/32264
http://build.chromium.org/p/tryserver.chromium/builders/mac/builds/32445
http://build.chromium.org/p/tryserver.chromium/builders/win/builds/37228

[James Robinson]

What about the layout tests trybots? (mac_layout, win_layout, linux_layout and the same with _rel for release).

[Sailesh Agrawal]

(In reply to comment #14)
> What about the layout tests trybots? (mac_layout, win_layout, linux_layout and the same with _rel for release).

Damn, forgot. Layout try bots pending.

[Sailesh Agrawal]

Hm.. a couple of webkit_gpu_tests failed. Not sure if it's my fault. I'm sending out a new layout try bot test with an empty change to check.

http://build.chromium.org/p/tryserver.chromium/builders/mac_layout/builds/667
http://build.chromium.org/p/tryserver.chromium/builders/mac_layout_rel/builds/205
http://build.chromium.org/p/tryserver.chromium/builders/win_layout/builds/948
http://build.chromium.org/p/tryserver.chromium/builders/win_layout_rel/builds/289
http://build.chromium.org/p/tryserver.chromium/builders/linux_layout/builds/794
http://build.chromium.org/p/tryserver.chromium/builders/linux_layout_rel/builds/270

[Sailesh Agrawal]

Yay, an empty change has the same failures:
http://build.chromium.org/p/tryserver.chromium/builders/mac_layout_rel/builds/207

[Mihai Parparita]

Comment on attachment 97369 [details]
Patch

Rubber-stamping since James is out.

[WebKit Review Bot]

Comment on attachment 97369 [details]
Patch

Clearing flags on attachment: 97369

Committed r89065: <http://trac.webkit.org/changeset/89065>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.