[Webkit-unassigned] [Bug 220115] [GTK][WPE] UserInteractive threads are not Real-time in Linux

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=220115

--- Comment #54 from Michael Catanzaro <mcatanzaro at gnome.org> ---
(In reply to Carlos Garcia Campos from comment #52)
> This is always called in the main thread.

That doesn't mean it's OK to use.

Anyway, we have bug #231913 for this. I suppose I can always just update the patch there if this one lands first.

> >> Source/WTF/wtf/linux/RealTimeThreads.cpp:173
> >> +    RELEASE_ASSERT_NOT_REACHED();
> > 
> > Woah! We assert if an rtkit properly is invalid. That's not programmer error in WebKit, and it's not enough reason to crash either. Instead, call g_set_error_literal() with some canned message and return -1.
> 
> That would mean it's not the rtkit service we expect to talk to, we always
> expect the properties to be either a int64 or int32.

That's not a WebKit programmer error, so assert is definitely inappropriate.

Crashing seems inappropriate too. What's wrong with returning a normal error.

> RunLoop::Timer is thread safe. it's a *main loop* timer, so the callback is
> always called in the main thread. And m_realTimeKitProxy is also called in
> the main thread only, because promoteThreadToRealTime() is always called in
> the main thread.

OK, can you add some ASSERT(isMainThread()) then? That would make the safety easier to see.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20211019/6f13872b/attachment.htm>