1849181 - Regression on AWFY-Jetstream2-Chai* tests around 15Aug2023

Status: RESOLVED WONTFIX

(Core :: JavaScript Engine: JIT, defect, P1)

Description

[Mayank Bansal]

See this : https://treeherder.mozilla.org/perfherder/graphs?highlightAlerts=1&highlightChangelogData=1&highlightCommonAlerts=0&series=mozilla-central,3737950,1,13&series=autoland,3911337,1,13&timerange=5184000&zoom=1691966466261,1692375424341,57.521907676272,81.19505303383397

Initial suspect: bug 1824051
If it actually is bug 1824051, it feels that the perf improvement from that bug are far more than this one regression.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

This bug has been marked as a regression. Setting status flag for Nightly to affected.

Comment 2

[Sam Foster [:sfoster] (he/him)]

Both the suspect regressors are in the JIT component, so lets move this bug there.

Comment 3

[Iain Ireland [:iain]]

Doug, can you take a quick look at this? I think we're happy to eat this regression in exchange for wins elsewhere, but maybe there's an easy fix.

Comment 4

[Alex Thayer [:alexical] (she/her)]

(In reply to Iain Ireland [:iain] from comment #3)

Doug, can you take a quick look at this? I think we're happy to eat this regression in exchange for wins elsewhere, but maybe there's an easy fix.

Yup!

Comment 5

[Alex Thayer [:alexical] (she/her)]

So, it looks like what's happening is they are constantly changing the trap function out from the handler. I'm not sure where in the framework this is happening, but we keep reattaching GetScriptedProxy with different trap functions for the same pc_. This feels like "testing frameworks gonna testing framework", and there may be something we could do about it, but it seems pretty low priority.

Comment 6

[Donal Meehan [:dmeehan]]

Setting Bug 1824051 as the regressor, please correct if wrong.
Keeping on the radar for 118 pending investigation in Comment 5

Comment 7

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1824051

Comment 8

[Iain Ireland [:iain]]

I agree that this seems low priority. I think it's best to close this. If we ever improve our handling of this case, it will be because we find code in the wild that benefits, not for a small improvement on a somewhat questionable sub-benchmark.