1991587 - First-paint-equals-lcp-text test on LCP Web Platform Tests is failing intermittently

Status: RESOLVED FIXED

(Core :: DOM: Performance APIs, defect, P3)

Description

[Justin Link]

The "first-paint-equals-lcp-text" LCP WPT test is currently failing with the assert " !EQ(64, 52)".

Comment 1

[Nazım Can Altınova [:canova]]

It looks like we don't use the same timestamp for the LCP and FCP entries. The current spec implements them in a way that they share the same timestamp. Writing a patch that makes this part more spec compliant.

Comment 2

[Nazım Can Altınova [:canova]]

Attachment: Bug 1991587 - Use consistent timestamp for FCP and LCP entries r?jlink,smaug

FCP and LCP were using different timestamps for the same paint. FCP used
RefreshDriver()->MostRecentRefresh() (tick start time) while LCP used
TimeStamp::Now() captured in the function, causing LCP to be reported
differently compared to FCP.

Both metrics now capture TimeStamp::Now() once and share it. This matches
the Paint Timing spec which says to capture "rendering update end time"
when mark paint timing runs, not reuse the tick start timestamp:
https://w3c.github.io/paint-timing/#mark-paint-timing

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/fe854316afde
https://hg.mozilla.org/integration/autoland/rev/f514773641ed
Use consistent timestamp for FCP and LCP entries r=jlink,smaug

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/746419bb1939
https://hg.mozilla.org/integration/autoland/rev/ab31d3b513d2
Revert "Bug 1991587 - Use consistent timestamp for FCP and LCP entries r=jlink,smaug" for causing wpt failures @fcp-ensure-update-the-rendering-step.html.

Comment 5

[agoloman]

Backed out for causing wpt failures @fcp-ensure-update-the-rendering-step.html.

• Backout link
• Push with failures
• Failure Log

Comment 6

[Nazım Can Altınova [:canova]]

It looks like the failing wpt test /paint-timing/fcp-only/fcp-ensure-update-the-rendering-step.html was never upstreamed:
The PR was closed before merging due to Chrome failures: https://github.com/web-platform-tests/wpt/pull/25578
And if you look at the wpt repo, there is no test with that name: https://github.com/search?q=repo%3Aweb-platform-tests%2Fwpt%20fcp-ensure-update-the-rendering-step.html&type=code

I also run the test in Chrome, and Chrome also fails consistently. This is because the test was actually asserting the old Firefox behavior that was not spec compliant. We should just remove that test instead.

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/0b1b8ae99ebb
https://hg.mozilla.org/integration/autoland/rev/c7c1f6d35b43
Use consistent timestamp for FCP and LCP entries r=jlink,smaug

Comment 8

[Cristina Horotan [:chorotan]]

https://hg.mozilla.org/mozilla-central/rev/c7c1f6d35b43

Comment 9

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Failed to create upstream wpt PR due to merge conflicts. This requires fixup from a wpt sync admin.