Perfherder has detected a awsy performance regression from push bbf90b97018afa7599012fb67787c0814cc2091b. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests. Please follow our guide to handling regression bugs and let us know your plans within 3 business days, or the offending patch(es) will be backed out in accordance with our regression policy.

For more information on performance sheriffing please see our FAQ.

(In reply to Andra Esanu from comment #0)

Regressions:

Ratio	Suite	Test	Platform	Options	Absolute values (old vs new)
2%	Base Content JS		macosx1014-64-shippable-qr		2,798,629.33 -> 2,841,002.67

We are aware that this might have caused a small regression across platforms because both the framescript and actor implementations are running at the same time, but with bug 1669169 all commands and events are getting routed through actors now. Next step for us is to not even register the framescript when actors are enabled. This work should be done end of this week or early next week (see bug 1669174).

What I don't understand is why the base content JS test only regresses on Mac but not on other platforms (Linux, Windows). Andra, do you have some links to these?

Nevertheless I would suggest that we ignore this warning, and wontfix this bug.

Henrik, the other platforms' Base Content JS tests went pretty noisy (linux1804-64-shippable, windows10-64-shippable-qr) and perfherder didn't detected them as regressions. Graph here. I wouldn't rush on WONTFIXing this bug regression given the noise it caused.

Ok, thanks! So lets keep this bug open and revisit it once bug 1669174 has been landed. It should happen later this week or early next week.

== Change summary for alert #27686 (as of Mon, 16 Nov 2020 14:18:36 GMT) ==

Regressions:

For up to date results, see: https://treeherder.mozilla.org/perfherder/alerts?id=27686

(In reply to Alexandru Ionescu (needinfo me) [:alexandrui] from comment #4)

== Change summary for alert #27686 (as of Mon, 16 Nov 2020 14:18:36 GMT) ==

Regressions:

Ratio	Suite	Test	Platform	Options	Absolute values (old vs new)
452%	reddit	fcp	android-hw-g5-7-0-arm7-api-16-shippable	warm webrender	283.75 -> 1,565.33
323%	reddit	FirstVisualChange	android-hw-g5-7-0-arm7-api-16-shippable	warm webrender	394.92 -> 1,669.58
249%	reddit	fcp	android-hw-p2-8-0-android-aarch64-shippable	warm webrender	206.21 -> 719.58
217%	reddit	FirstVisualChange	android-hw-p2-8-0-android-aarch64-shippable	warm webrender	240.42 -> 761.08
119%	reddit		android-hw-g5-7-0-arm7-api-16-shippable	warm webrender	764.20 -> 1,672.30

All those so called regressions in Browsertime jobs aren't tied to Marionette and the enabling of actors. When you check the graphs for each of those individual regressions you can see that numbers dropped again on Nov 11th:

https://treeherder.mozilla.org/perfherder/graphs?timerange=1209600&series=autoland,2857959,1,13

I haven't checked them all, but I highly suspect that all these regressions are unrelated here.

Henrik, seems like a bunch of alert items in the summary were not caused by your patch. Thus, this alert summary contains only 2 items that are improvements. Closing this as invalid. Thanks for your input.