performance.clearMeasures (and presumably performance.clearMarks) results in inconsistency between LH and other tools

[andydavies]

FAQ

• Yes, my issue is not about variability or throttling.
• Yes, my issue is not about a specific accessibility audit (file with axe-core instead).

URL

https://andydavies.github.io/agent-tests/usertiming/clearedmeasures.html

What happened?

The page sets some User Timing marks and a measure, and then clears the measure after a delay

WebPageTest (https://www.webpagetest.org/result/240112_AiDcBH_827/), other tools don't report the measure as they're no longer in the Performance Timeline at the point WPT captures the data.

LH extracts the marks & measures from the Chrome trace and has no awareness of calls to clearMarks / clearMeasures so includes marks and measures that may have been removed

What did you expect?

I'd expect LH to not include marks / measures that were removed but I can see the case for excluding them

The discrepancy does cause confusion (we have a SpeedCurve customer who raised this initially)

I'm OK if you want to close this as WAI but just wanted to document the issue for the next person who discovers this difference

What have you tried?

No response

How were you running Lighthouse?

WebPageTest

Lighthouse Version

11.4.0

Chrome Version

No response

Node Version

No response

OS

No response

Relevant log output

No response

[adamraine]

My initial thoughts are that this is WAI since we want the audit to document everything that happened. Performance panel in DevTools also appears to keep the performance timings visible.

Is there a use case for clearing the timings or was it just confusion as a result of differences with WPT?