The meta file for the given test still contains expected fails for the test test_release_mouse_sequence_resets_dblclick_state. Those have been added because an element origin is used there and it was not supported before the patch on bug 1806788 had been landed.

Now those entries shouldn't be there anymore given that this bug has been fixed in Firefox 114. What me makes wonder now is why we do not see the unexpected pass failures at all on autoland and mozilla-central, but this try build.

Ah, so the upstream fix for the tests as synced via bug 1834202 actually fixed that. But why did the sync bot not notice that as well?

James do you have an idea? I actually don't know.

https://treeherder.mozilla.org/jobs?repo=mozilla-central&selectedTaskRun=WKPPp2nUSjyvbJ6UNrx4tw.0&searchStr=android%2Cweb-platform-tests suggests we aren't running the fission android tests on mozilla-central. That might explain why we didn't have that metadata, although I don't know why we've got builds on try that aren't on central.

Thanks. Yeah, that is very suspicious. Maybe Joel knows more about those default configurations for try but he seems to be out right now. Marco, who could best help out here instead? Thanks.

Maybe ahal, aryx or gbrown can help, otherwise we can wait for Joel to be back.

web-platform wdspec tasks are configured to not run with the default configuration (= fission) on Android.

See also bug 1714654 which tracks fixing and enabling web-platform tests with Fission on Android. Should wdspec run for different configurations than wpt?

Thanks Aryx! I don't know why we do not run the tests with Fission on Android when it is the default config. As best for webdriver we should run what we actually ship to users. Chris, what do we actually ship and if it's with Fission enabled, why don't we run the wdspec tests under Fission? Thanks.

We're not shipping Fission on Android at this time, so Fission should not be the default test config on Android. We're working on Fission support, so we do run a subset of Android tests in both Fission and non-Fission (e10s) modes, if they are known to pass. I'll make sure wdspec tests are on our list of Android tests to make Fission-compatible. I estimate we will ship Android Fission in 2023 Q4.

Does that explain what you are seeing? Do we need to change the wdspec test config for Android? Does this prevent you from removing test meta data for this bug?

https://hg.mozilla.org/mozilla-central/rev/af393b66d77c

(In reply to Chris Peterson [:cpeterson] from comment #9)

Does that explain what you are seeing? Do we need to change the wdspec test config for Android? Does this prevent you from removing test meta data for this bug?

Thanks for the details. It actually doesn't fully explain the difference in test selection (fission vs. non fission) between try and integration branches to me, but overall we should really make sure to run the wdspec tests in both modes. Some tests are still failing and we definitely have to track and fix those before Firefox on mobile gets released with Fission enabled.