It's possible for the video provider and frame to go away after willDraw, but before the frame gets uploaded into textures. Since the frame is a raw pointer (that the provider is responsible for), this could cause a bad memory access. Unfortunately, this depends on some other fixes to make sure that we don't ever call willDraw a second time without calling didDraw (which will deadlock on aquiring the exact same lock).
Created attachment 138712 [details] Patch
Comment on attachment 138712 [details] Patch R=me
Committed r115281: <http://trac.webkit.org/changeset/115281>