1710549 - Refine the way to tell whether we should disable APZ for subframes during using devtools

Status: ASSIGNED

(Core :: Panning and Zooming, defect, P3)

Description

[Hiroyuki Ikezoe (:hiro)]

As botond noticed we disable APZ if there's any anonymous contents having a valid nsIFrame checked in HasVisibleAnonymousContents which is a situation when devtools is open and it's expected when devtools' highligther is visible.

But in reality, the anonymous content is this class=highlighter-container div and it has an nsIFrame so unfortunately the function HasVisibleAnonymousContents doesn't work for the devtools case as expected. What element devtools tries to hide/show is this child element instead. This is quite confusing when testing overscrolling with devtools. Overscrolling which is handled by APZ doesn't work during opening devtools.

I wonder why the top anonymous content needs to be in terms of layout (i.e. needs to have the nsIFrame).

Comment 1

[Hiroyuki Ikezoe (:hiro)]

It looks like we initially wanted to set display:none to the anonymous content (display:none destroys the nsIFrame and all descendant nsIFrames), but it's not been working as expected?

Gabriel, any insight on this?

Comment 2

[Gabriel Luong [:gl] (ΦωΦ)]

Sorry, I do not have any insight on this.

Comment 3

[Emilio Cobos Álvarez (:emilio)]

I actually suspect this code no longer works as expected. The host now has display: contents here so it never has a frame.

I guess AccessibleCaret still has a frame tho, but do we want to disable APZ there?

In any case I think this is either WFM or WONTFIX or that code should be removed. Thoughts hiro?

Comment 4

[Hiroyuki Ikezoe (:hiro)]

At least on my Linux box, I don't see highlighters are behind from its proper scroll position even if I forcibly enable APZ in devtools.

So probably we can drop the code. Let's see if there's any test affected by the drop.
https://treeherder.mozilla.org/jobs?repo=try&revision=201c850143c31e851b5f6d95f904f738843dfe1f

Comment 5

[Hiroyuki Ikezoe (:hiro)]

I am supposing bug 1730998 and bug 1571758 contributed the original highlighter issue.