1725541 - Make test_layerization.html work on xorig runs

Status: RESOLVED FIXED

(Core :: Panning and Zooming, task, P2)

Description

[Hiroyuki Ikezoe (:hiro)]

It still doesn't work with the changes for bug 1725540. The errors are mostly same as bug 1722907 comment 0;

In the xorigin case there's an additional error;

inner4 still is directly activated because it never expired

Comment 1

[Chris Peterson [:cpeterson]]

Hiro, does this test_layerization.html test fail in xorigin mode regardless of whether Fission is enabled or not? Do we need to fix this test for Fission?

Comment 2

[Hiroyuki Ikezoe (:hiro)]

I just run it locally to make sure the situation. Actually the test also fails with disabling Fission, but in a different way. It causes a time out rather than the errors in comment 0.

Comment 3

[Frank Griffith Jr]

Assigned to Tim. Linking to MVP for now. If found not urgent can push to future

Comment 4

[Chris Peterson [:cpeterson]]

This test is currently skipped on Android, headless Fission Linux, and xorigin mode:

https://searchfox.org/mozilla-central/rev/aa46c2dcccbc6fd4265edca05d3d00cccdfc97b9/gfx/layers/apz/test/mochitest/mochitest.ini#62-66

[test_layerization.html]
  skip-if =
    os == 'android'  # wheel events not supported on mobile
    os == 'linux' && fission && headless # Bug 1722907
    xorigin # "scrolling 'inner2' should cause it to be directly activated" etc.

Bug 1722907 is another test_layerization.html test failure.

Comment 5

[Timothy Nikkel (:tnikkel)]

Just to update after investigation I don't think there is a problem in the browser that we want to fix to make this test pass. There are some differences in the size of display ports that happen when we have process boundaries vs when we don't, and the test was unintentionally relying on having the sizes without process boundaries.

The test is run inside the usual mochitest iframe, meaning half of the content is not visible, and since displayports are based on what is visible, and the test checks against the a multiple of the full element height, the displayports height is not enough when we have a process boundary.

The fix would be one of the following:

1. use a multiple of the visible element height to compare the displayport height against
2. shrink the content so it fits in the mochitest iframe
3. open the test in a new tab so the full content is visible, and maybe put it in an xorigin iframe in the new tab so we still test the xorigin case

Since there isn't a code fix here and only a test fix this doesn't really need to block mvp.

Comment 6

[Chris Peterson [:cpeterson]]

(In reply to Timothy Nikkel (:tnikkel) from comment #5)

Since there isn't a code fix here and only a test fix this doesn't really need to block mvp.

In that case, I'll move this bug to from Fission Milestone MVP to our Future backlog. Thanks for the explanation!

Comment 7

[Hiroyuki Ikezoe (:hiro)]

Attachment: Bug 1725541 - Consider that the test runs inside an iframe whose size is smaller than the target element's size. r?tnikkel

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/ece3dcc343a2
Consider that the test runs inside an iframe whose size is smaller than the target element's size. r=tnikkel

Comment 9

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/ece3dcc343a2

Comment 10

[Chris Peterson [:cpeterson]]

Setting status-firefox93=wontfix because we don't need to uplift this test fix to Beta 93.