Window occlusion caused test timeout failures on several test. It seems necessary to disable window occlusion on these tests.
https://treeherder.mozilla.org/jobs?repo=try&revision=16c4bc573147b78f232926b3cd0476986c51770e

See Bug 1688997 comment 19

Depends on D126710

From gfx-apac meeting

• It is OK to disable Window Occlusion on web platform test
• We need to run common test suites with window occlusion.

Test results without disabling window occlusion, there are many timeout failures.
https://treeherder.mozilla.org/jobs?repo=try&revision=7a63a269ffe34c6aad407cab4a517b8ed27325e1

With D127446, almost all timeout failure seemed to be addressed.
https://treeherder.mozilla.org/jobs?repo=try&revision=458c7ecb0a8494d06574835ef9427c08854c878f

Looking into why testing window was occluded by other window during test.
There were 3 causes to cause window occlusion

• [1] Multiple Firefox instances existed. A window of one instance occluded other window of another instance.
  • This happened frequently during web platform tests.
• [2] One test created multiple windows. One window occluded other window.
  • On CI, display size seemed small, it caused testing windows were often maximized.
  • Multiple maximized windows caused window occlusion.
• [3] window is not shown(does not have WS_VISIBLE). The window is judged as occluded.

Looked into why tests timeout failures by window occlusion did not happen on macOS.

• On macOS, "occlusionstatechange" event happened less often. It seems to related to the failures difference.
• On macOS, window's sizes were not maximized during tests. On Windows, Majority of windows were maximized.

Maximized windows caused "occlusionstatechange" events on Windows.
It might be related to window size difference. macOS had bigger display size.

• macOS display size: 1920 * 1080
• Windows display size: 1280 * 1024

When non-maximized Firefox windows had same size and position, it caused "occlusionstatechange" event on Window.
But it did not cause "occlusionstatechange" event on macOS.
On macOS, window corners are rounded. I wonder if it might be related to it.

• https://treeherder.mozilla.org/jobs?repo=try&revision=7d7a489dea0e60fd0071184f10b6b674f6bad23c
• https://treeherder.mozilla.org/jobs?repo=try&revision=bf89ba55d72d1815c6846d473597c699dc011da0

Looked into why testing windows were maximized on Windows.

CI testing PC's display size was (1280 * 1024). With the size, following code was hit. It caused the maximized window.

• https://searchfox.org/mozilla-central/rev/483cfde5a54f6c1cd94c6295564993aeb4f10980/browser/base/content/browser.js#1620

Test result with patches.

• https://treeherder.mozilla.org/jobs?repo=try&revision=205c3be27978c986874d76ca198fa9d81a991810

test result without change.

• https://treeherder.mozilla.org/jobs?repo=try&revision=40ae66d61568de9e78fea136fb2643701dfa16e9

https://hg.mozilla.org/mozilla-central/rev/ae7da4eda98f