On 24.09.15 16:13, Alexander Scherbatiy wrote:
Right, it overrides the sendEvent and for some subtype="ExecuteBlockEvent" it tries to call a block. It means that if LWCToolkit.nextEventMatchingMask gets an event which is related to a deallocation it will later call release(which is inside the block) in the sendEvent. And the previous fix was an attempt to eliminate this situation of deallocation inside the nested loop.Then I do not have an idea why the crash has been fixed. May be an author of the fix 8068886 or reviewers can provide the real explanation.
The crash was fixed because the nativeCFRelease() now use events instead of selectors for resources deallocation. And even with an incorrect mask in the nextEventMatchingMask(), we work properly, because in most of the time we start our nested event loop with processEvents=false. Which means that we simply block all events including events for deallocation. But the crash still be possible in case of dnd where we proceed the events.
The fix looks fine to me. -- Best regards, Sergey.
