[Cocoa] [GTK] RunLoop::Timer::isActive() is incorrect for timers while they are firing

[Geoffrey Garen]

RunLoop::Timer::isActive() is true for one-shot timers even when they're not active

[Geoffrey Garen]

Created attachment 403168 [details]
Patch

[Geoffrey Garen]

Looks like a variant of gtk had this bug too.

[Geoffrey Garen]

Created attachment 403223 [details]
Patch

[Darin Adler]

Comment on attachment 403223 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=403223&action=review

> Source/WTF/wtf/cf/RunLoopCF.cpp:95
> +    if (!CFRunLoopTimerDoesRepeat(cfTimer))
> +        timer->stop();
> +
>      timer->fired();

Could we have put this code inside TimerBase::fired instead? Or does it not have an efficient way to check if the timer is non-repeating?

If stop() would do the trick on GLib too that would be neat.

[Geoffrey Garen]

Glib already had the implicit stop behavior. It needed the inverse fix: be sure to restart if you are a repeating timer. 

I do think we can/should do a shared implementation of isActive in TimerBase. Windows already has all the code we need; just need to migrate it. I hope to return and do that after I finish off callOnMainThread.

[EWS]

Committed r263783: <https://trac.webkit.org/changeset/263783>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 403223 [details].

[Radar WebKit Bug Importer]

<rdar://problem/64967584>

[Geoffrey Garen]

Reopening to attach new patch.

[Geoffrey Garen]

Created attachment 403265 [details]
Assertion fix

[EWS]

Committed r263786: <https://trac.webkit.org/changeset/263786>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 403265 [details].