Bug 201767 – REGRESSION (iOS 13): rAF stops firing when navigating away cross-origin and then back

• Component: Animations
• Status: RESOLVED FIXED

[Ali Juma]

Created attachment 378740 [details]
Test case

Steps to reproduce (from https://bugs.chromium.org/p/chromium/issues/detail?id=1003803):
1) Visit the attached test page. Notice the counter gets updated once every 10 frames using rAF.
2) Navigate away, either by clicking on the button or just by typing a URL in the usual way.
3) Go back to the first page.

Expected result:
The counter resumes updating.

Actual result:
The counter is either frozen or updates very slowly (e.g., only updates while pinch-zooming and then stops again).

This seems to be related to the fact the test case only updates the page once every 10 rAF callbacks; if I change this to update the page on every rAF, the bug no longer repros.

[Ali Juma]

Created attachment 378742 [details]
Test case

It also seems important that the URL used in Step 2 is cross-origin wrt the URL from step 1.

[Simon Fraser (smfr)]

That suggests that it's PSON-related.

[Radar WebKit Bug Importer]

<rdar://problem/55350854>

[Antoine Quint]

Can you confirm this is with an iOS 13.1 beta? This is what I can gather from the Chromium bug but it would be nice for this bug to contain the configuration.

[Ali Juma]

(In reply to Antoine Quint from comment #4)
> Can you confirm this is with an iOS 13.1 beta? This is what I can gather
> from the Chromium bug but it would be nice for this bug to contain the
> configuration.

This reproduces in iOS 13.1 beta 3, but also reproduces in iOS 13.0. However, it does not reproduce in iOS 12.4.

[Ali Juma]

Debugging this a bit, we're stuck in a state where RenderingUpdateScheduler::isScheduled is true, but RenderingUpdateScheduler::displayRefreshFired never gets called.

[Chris Dumez]

Ok, I can reproduce with the test case attached, thanks. I will investigate.
It does not likely related to PSON although, PSON really did not change much in iOS 13.

[Chris Dumez]

I have a fix, working on an automated test.

[Chris Dumez]

Created attachment 378904 [details]
Patch

[Geoffrey Garen]

Comment on attachment 378904 [details]
Patch

r=me

[EWS Watchlist]

Comment on attachment 378904 [details]
Patch

Attachment 378904 [details] did not pass win-ews (win):
Output: https://webkit-queues.webkit.org/results/13037841

New failing tests:
http/tests/navigation/page-cache-requestAnimationFrame.html

[EWS Watchlist]

Created attachment 378918 [details]
Archive of layout-test-results from ews214 for win-future

The attached test failures were seen while running run-webkit-tests on the win-ews.
Bot: ews214  Port: win-future  Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

[Chris Dumez]

Comment on attachment 378904 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=378904&action=review

> LayoutTests/http/tests/navigation/page-cache-requestAnimationFrame.html:1
> +<!-- webkit-test-runner [ enablePageCache=true ] -->

I need to add support for this to DRT so the test passes for WebKit1. Will do so before landing.

[Chris Dumez]

Created attachment 378966 [details]
Patch

[Chris Dumez]

Comment on attachment 378966 [details]
Patch

Clearing flags on attachment: 378966

Committed r249961: <https://trac.webkit.org/changeset/249961>

[Chris Dumez]

All reviewed patches have been landed.  Closing bug.

[Chris Dumez]

Reopening for follow-up since this introduced crashes.

[Chris Dumez]

Created attachment 379276 [details]
Patch

[Tim Horton]

Comment on attachment 379276 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=379276&action=review

> LayoutTests/http/tests/navigation/page-cache-requestAnimationFrame.html:42
> +        }, 100);

This is ... not ideal, but OK

[Chris Dumez]

Comment on attachment 379276 [details]
Patch

Clearing flags on attachment: 379276

Committed r250165: <https://trac.webkit.org/changeset/250165>

[Chris Dumez]

All reviewed patches have been landed.  Closing bug.