Bug 167832 – Fix misleading comment in RunLoop.h

• Status: REOPENED
• Component: Web Template Framework

[Joseph Pecoraro]

Fix misleading comment in RunLoop.h.

Mac initialization used to force using CFRunLoopGetMain(). Now however it just uses RunLoop::current which uses CFRunLoopGetCurrent(). So this comment that it can be done on any thread is misleading and can lead to incorrect behavior if it is actually done on a non-main thread on Mac.

[Joseph Pecoraro]

Created attachment 300604 [details]
[PATCH] Proposed Fix

[Alexey Proskuryakov]

Comment on attachment 300604 [details]
[PATCH] Proposed Fix

View in context: https://bugs.webkit.org/attachment.cgi?id=300604&action=review

> Source/WTF/wtf/RunLoop.h:53
> +    // Must be called from the main thread.

On the Mac, this can be checked with an assertion (using pthread_main_np).

[Joseph Pecoraro]

Created attachment 300643 [details]
[PATCH] Proposed Fix

[Joseph Pecoraro]

Comment on attachment 300643 [details]
[PATCH] Proposed Fix

Oh, that makes sense. WebKit1 will need some kind of "wait a beat".

[Joseph Pecoraro]

Comment on attachment 300643 [details]
[PATCH] Proposed Fix

That comment was meant for another bug.

[WebKit Commit Bot]

Comment on attachment 300643 [details]
[PATCH] Proposed Fix

Clearing flags on attachment: 300643

Committed r212139: <http://trac.webkit.org/changeset/212139>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[WebKit Commit Bot]

Re-opened since this is blocked by bug 168152

[Joseph Pecoraro]

This revealed a few issues where non-main thread code triggers WebKit2Initialize and so could potentially be initializing the main run loop on a non-main thread (because WebKit2Initialize is not dispatch_once and always calls down into RunLoop::initializeMainThread).

For example, this case in DumpRenderTree: (Bug 168149)
https://bugs.webkit.org/show_bug.cgi?id=168149