Bug 155130 – Fix lifetime issues regarding WebVideoFullscreenInterfaceMac

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit2

[Ada Chan]

When WebVideoFullscreenManager::setUpVideoControlsManager() is called, we always clean up the old WebVideoFullscreenInterfaceContext associated with the context id associated with the current video controls manager and make a new one.  This is an issue if we are in the middle of handling fullscreen and it's still using that WebVideoFullscreenInterfaceContext, since a lot of fullscreen state is kept with that particular context.

Before the introduction of video controls manager, we only use WebVideoFullscreenInterfaceContext for full screen, so it's ok to invalidate the interface right after we exit the fullscreen mode. But now we need to do more bookkeeping since video controls manager also depends on WebVideoFullscreenInterfaceContext. We should keep track of a "client count" for each context ID so we'll only remove it from the context map when there are all clients are done with it.

[Ada Chan]

<rdar://problem/24978679>

[Ada Chan]

Created attachment 273209 [details]
Patch

[WebKit Commit Bot]

Attachment 273209 [details] did not pass style-queue:


ERROR: Source/WebKit2/WebProcess/cocoa/WebVideoFullscreenManager.mm:336:  More than one command on the same line  [whitespace/newline] [4]
Total errors found: 1 in 5 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Beth Dakin]

This looks good to me. It would be nice to get confirmation fro Jer or Eric that this seems okay.

[Jer Noble]

Comment on attachment 273209 [details]
Patch

This is a little weird, since we're essentially reproducing RefCounted.  It seems like we should be able to do this just through the reference count of the final Model and Interface objects.  But we can revisit that idea later, and maybe even use this m_clientCount concept to do so. LGTM.

[Ada Chan]

(In reply to comment #5)
> Comment on attachment 273209 [details]
> Patch
> 
> This is a little weird, since we're essentially reproducing RefCounted.  It
> seems like we should be able to do this just through the reference count of
> the final Model and Interface objects.  But we can revisit that idea later,
> and maybe even use this m_clientCount concept to do so. LGTM.

Yep, agreed.  

Committed:
http://trac.webkit.org/changeset/197787