There's a slight performance hit here by not running them concurrently (now takes ~2.5s on Chrome and Firefox on CI, versus 1 and 1.8s before). If we care about that, we can make these properly async_tests by not using await (they don't actually need it, but this was the easy fix).

Weirdly, there's a diff in Safari that I haven't figured out yet - https://wpt.fyi/results/html/browsers/windows/clear-window-name.https.html?diff&filter=ADC&run_id=688120008&run_id=669410010 . This test may be flaky on Safari (it was just added, and doesn't have enough wpt.fyi runs to figure that out), or I may have inadvertently triggered some sort of behavior change. Perhaps you can try running it locally @foolip to see if you can reproduce with/without?

Sorry for missing this in review. This looks reasonable to me. @artines1 might have thoughts as the original author.

Sorry for missing this in review. This looks reasonable to me. @artines1 might have thoughts as the original author.

No worries, I assigned to foolip anyway ;)

I'll gives artines1 24 hours to see if they have any comments, then I'll merge this.

The patch looks good to me. I don't have any comments. Thanks.

@stephenmcgruer I've tried running ./wpt run --no-pause --channel=preview safari html/browsers/windows/clear-window-name.https.html at commit e0020b8 (without this change) and subtest "Window.name is set by the window" times out every time. It takes 60 seconds, so I only tried 3 times.

@foolip then it seems like I have introduced a behavior change (probably by making it non-concurrent), but unclear to me what. If you're able to do any debugging that would be great, there may be a Safari bug hiding here? (Something about simultaneous windows being opened...?)

@stephenmcgruer when running the tests with your change, it all happens very fast. Windows open and close very fast and then the test is done. Before, one window was just left sitting open.

I've not been able to get to the bottom of it, but it looks like what matters is what happens after the hanging test. By removing all the prior tests (leaving just the last two) and leaving just the call to anchorClick in the second test, the hang still repros. The problem isn't in pollResultAndCheck, but something about clicking to open another window after "Window.name is set by the window" breaks the test. It doesn't seem to matter what URL is opened in the second test.

I'd need to spend more time minifying this, more time than I have today.

@foolip - we still have the surprising improvement in Safari that I don't understand here. Are you still happy to merge this?

Safari reuses the top-level window when you use <a>.click(). That explains the difference. That doesn't really seem non-conforming so this seems like a better way to test this.

Safari reuses the top-level window when you use .click(). That explains the difference. That doesn't really seem non-conforming so this seems like a better way to test this.

Ack, thanks!

Ah, thanks for explaining what's going on @annevk!