Bug 219052 – [macOS] Stop using RunLoopType=_WebKit starting in Big Sur

• Component: WebKit2
• Status: RESOLVED FIXED

[Chris Dumez]

Stop using RunLoopType=_WebKit starting in Big Sur. This was a temporary hack so that our WebProcesses would get the right scheduling priority. Starting with Big Sur, the right way to do this is to use _ProcessType=App and restore the RunLoopType to be NSRunLoop.

[Chris Dumez]

<rdar://problem/61742969>

[Chris Dumez]

Created attachment 414383 [details]
Patch

[Geoffrey Garen]

It looks like r233932 was about app nap / tab nap. With this change, do tabs still nap / suspend properly?

[Geoffrey Garen]

I have a vague memory that we made a sandboxing change that was incompatible with the App process type or the NSRunLoop runloop type. Maybe Per Arne remembers?

[Chris Dumez]

(In reply to Geoffrey Garen from comment #4)
> I have a vague memory that we made a sandboxing change that was incompatible
> with the App process type or the NSRunLoop runloop type. Maybe Per Arne
> remembers?

I will verify AppNap shortly. I don't foresee any issue.

[Chris Dumez]

(In reply to Geoffrey Garen from comment #4)
> I have a vague memory that we made a sandboxing change that was incompatible
> with the App process type or the NSRunLoop runloop type. Maybe Per Arne
> remembers?

The sandboxing change made it so that we could not use _NSApplicationMain RunLoopType anymore. We initially tried to use NSRunLoop RunLoopType instead but it broke AppNap. This is when the _WebKit RunLoopType was introduced to unbreak AppNap. I am being told that we can now revert to NSRunLoop RunLoopType as long as we use "App" as _ProcessType (See radar).

[Chris Dumez]

Created attachment 414385 [details]
Patch

[Chris Dumez]

Comment on attachment 414385 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=414385&action=review

> Source/WebKit/ChangeLog:13
> +        After this change, I have verified that the WebContent's main thread still runs at UserInteractive QoS. App Nap is also

I have verified that AppNap is still working fine with this patch.

[Geoffrey Garen]

Comment on attachment 414385 [details]
Patch

r=me

OK, I think that addresses all my vague memories :P

[EWS]

Committed r269928: <https://trac.webkit.org/changeset/269928>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 414385 [details].

[Per Arne Vollan]

(In reply to Geoffrey Garen from comment #4)
> I have a vague memory that we made a sandboxing change that was incompatible
> with the App process type or the NSRunLoop runloop type. Maybe Per Arne
> remembers?

If I recall correctly, we stopped using the NSApplication runloop when blocking the WindowServer. As Chris noted, this introduced a process priority regression, which was fixed by using this custom runloop type. It was possibly also needed to fix the AppNap regression, but Chris seems to have that covered :)