1691358 - Desktop zoom value is doubled (or incorrect) in BrowserChild::mChildToParentConversionMatrix after PresShell::SetResolution call until layers get updated or RequestContentRepaint happens

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Hiroyuki Ikezoe (:hiro)]

During writing mochitest (with nsIDOMWindowUtils.setResolutionAndScaleTo) for bug 1684795, I noticed this issue.

What I wrote initially is calling setResolutionAndScaleTo after the target iframe has been loaded, then calling a pair of promiseApzFlushedRepaints() and waitUntilApzStable(), but with this setup, I couldn't get the correct result via BrowserChild::mChildToParentConversionMatrix.

When we compute the transform matrix, we call HitTestingTreeNode::GetTransformToGecko which calls AsyncPanZoomController::GetTransformToLastDispatchedPaint and in the function we do compare mExpectedGeckoMatrix.GetZoom() and mLastContentPaintMetric.GetZoom(), the code is;

    zoomChange = lastContentZoom / lastDispatchedZoom;

So, for example, if there is a SetResolution call with 2.0, then the zoomChange will be 2.0, then it will be cumulated with mTransform value in HitTestingTreeNode::GetTransformToGecko.

Comment 1

[Botond Ballo [:botond]]

I'm happy to investigate this a bit, but I'm unclear on how to repro it at the moment. Hiro, you mentioned a patch to browser_test_select_popup_position.js which I applied -- should I be seeing a problematic TransformToGecko when running the test?

Comment 2

[Hiroyuki Ikezoe (:hiro)]

Yeah, I think it should happen by running the test with --enable-fission option.

Comment 3

[Hiroyuki Ikezoe (:hiro)]

Also, the patch adds a setTimeout call so that during opening the content in question, you can click the <select> element to see where the popup window is shown.

Comment 4

[Timothy Nikkel (:tnikkel)]

I think I may have run into this again in bug 1719406.

Comment 5

[Botond Ballo [:botond]]

Do you mean that bug 1719406 fixes this?

Comment 6

[Timothy Nikkel (:tnikkel)]

I kinda think so, this bug sounds very very similar to what I was seeing, but I didn't test the original reproduction of this bug to be sure.

Comment 7

[Botond Ballo [:botond]]

Hiro, are you able to confirm whether this has been fixed?

Comment 8

[Hiroyuki Ikezoe (:hiro)]

Yep, it's been fixed. I am going to re-use this bug to add a test.

A try run with the test; https://treeherder.mozilla.org/jobs?repo=try&revision=a16d46a428426fcadf5729f3d085f254d87ba643

Comment 9

[Hiroyuki Ikezoe (:hiro)]

Attachment: Bug 1691358 - Add a test to make sure nsIDOMWindowUtils.setResolutionAndScaleTo properly delivered to OOP iframes. r?botond

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/2997688aff64
Add a test to make sure nsIDOMWindowUtils.setResolutionAndScaleTo properly delivered to OOP iframes. r=botond

Comment 11

[Alexandru Michis [:malexandru]]

https://hg.mozilla.org/mozilla-central/rev/2997688aff64