1702988 - [macOS] Elastic Overscroll doesn't work when mousecursor is over Adclick Ads

Status: RESOLVED DUPLICATE of bug 1705280

(Core :: Panning and Zooming, defect)

Description

[Mehmet]

Attachment: Bildschirmaufnahme 2021-04-05 um 15.31.44.mov

Nightly 89.0a1 (2021-04-05) (64-Bit)
macOS 11.2.3

1.) Be on macOS and set apz.overscroll.enabled to true.
2.) Visit a page with Adclick Ads e.g. https://9to5mac.com/
3.) Move the cursor over the Adclick Ad
4.) Scroll up to receive the Overscroll effect

Actual: The Overscroll Effect doesn't work
Expected: The Overscroll Effect should work

A screencast is attached. It works when I am outside the VMWare Ad. It doesn't work when I am inside the VMWare Ad.

Thanks.

Comment 1

[Botond Ballo [:botond]]

I can't reproduce this on Linux.

Comment 2

[Hiroyuki Ikezoe (:hiro)]

I can reproduce on Mac with enabling fission. That means OverscrollHandoff fails somewhere?

Comment 3

[Botond Ballo [:botond]]

Interesting. On Linux, I can't repro with Fission enabled, either.

Comment 4

[Hiroyuki Ikezoe (:hiro)]

What I've noticed is this issue depends on the Ads, I found with an Ad I am able to reproduce, but with others I am not.

Comment 5

[Mehmet]

(In reply to Hiroyuki Ikezoe (:hiro) from comment #4)

What I've noticed is this issue depends on the Ads, I found with an Ad I am able to reproduce, but with others I am not.

Yes, that is exactly the same behavior I noticed on macOS too. On some ads it fails and in some ads it works.

(BTW: What is fission? And it is then really related to?)

Thanks.

Comment 6

[Hiroyuki Ikezoe (:hiro)]

Fission is a project name to do site isolation, you can find/enable/disable it in "Nightly Experiments" in about:preferences.

Comment 7

[Mehmet]

(In reply to Hiroyuki Ikezoe (:hiro) from comment #6)

Fission is a project name to do site isolation, you can find/enable/disable it in "Nightly Experiments" in about:preferences.

Thanks for your feedback. I can reproduce with and without fission enabled with the same ad. So probably not fission related. Thanks.

Comment 8

[Chris Peterson [:cpeterson]]

(In reply to Mehmet from comment #7)

I can reproduce with and without fission enabled with the same ad. So probably not fission related. Thanks.

In that case, I will remove the "Fission" keyword from this bug.

Thanks for testing!

Comment 9

[Hiroyuki Ikezoe (:hiro)]

The issue I am seeing is definitely related to Fission.

A very easy way to reproduce this issue I found now is;

1. Make sure Fission enabled
2. Make sure apz.overscroll.enabled is true
3. Open https://hsivonen.fi/fission-host.html
4. Try to scroll up (try to overscroll) on an iframe there

Comment 10

[Mehmet]

(In reply to Hiroyuki Ikezoe (:hiro) from comment #9)

The issue I am seeing is definitely related to Fission.

A very easy way to reproduce this issue I found now is;

1. Make sure Fission enabled
2. Make sure apz.overscroll.enabled is true
3. Open https://hsivonen.fi/fission-host.html
4. Try to scroll up (try to overscroll) on an iframe there

Thanks, yes, I can confirm it:

Fission enabled: Overscroll fails at https://hsivonen.fi/fission-host.html
Fission disabled: Overscroll works at https://hsivonen.fi/fission-host.html

Comment 11

[Hiroyuki Ikezoe (:hiro)]

Attachment: A possible minimized test case for non fission

Though I haven't reproduced the issue on 9to5mac.com without Fission, this is probably a minimized test case which is able to reproduce the issue without Fission.

You can't vertically overscroll the root element on red area. And interestingly, it's not horizontally overscrollable in the red area.

I think this kind of cases commonly happen, that said, to fix this issue we will probably need to have the dominant axis scrolling, and even if the issue happens, users can scroll normally, and the issue is not so severe than issues that overscroll gets stuck, so I'd suggest this is an issue nice to have in MR1.

Comment 12

[Hiroyuki Ikezoe (:hiro)]

Just confirmed that the following change fixes the 9to5mac case (overflow:hidden on both directions) I've been seeing, also fixes Henri's fission test case (unscrollable on both directions).

--- a/gfx/layers/apz/src/AsyncPanZoomController.cpp
+++ b/gfx/layers/apz/src/AsyncPanZoomController.cpp
@@ -3442,7 +3442,7 @@ bool AsyncPanZoomController::AttemptScro
     return true;
   }
-
-  if (AllowScrollHandoffInCurrentBlock()) {
+  if (!IsPannable() || AllowScrollHandoffInCurrentBlock()) {
     // If there is overscroll, first try to hand it off to an APZC later
     // in the handoff chain to consume (either as a normal scroll or as
     // overscroll).

Probably we should fix these cases first, and defer the case in comment 11 in a later bug.

Comment 13

[Mehmet]

Thank you :)

Comment 14

[Hiroyuki Ikezoe (:hiro)]

Hi Mehmet, a relevant issue (bug 1705280) has been fixed in nightly, can you please try it to see if the issue you are seeing is still reproducible or not? Thanks!

Comment 15

[Mehmet]

(In reply to Hiroyuki Ikezoe (:hiro) from comment #14)

Hi Mehmet, a relevant issue (bug 1705280) has been fixed in nightly, can you please try it to see if the issue you are seeing is still reproducible or not? Thanks!

Hi Hiroyuki,

I retested it and it works for me now :)

Please feel free to close this report.

Many thanks
Mehmet

Comment 16

[Hiroyuki Ikezoe (:hiro)]

Thank you, Mehmet!