Review request for MACOSX_PORT-686: Regression in drawing (Not running on AppKit thread 0...)

[Dmitry Cherepanov]

Hi Anthony,

Thanks for looking into this.

I suppose that it's not related to JNFJObjectWrapper as Scott wrote the 
following comment in the bug report:

"The method is being called in a cleanup routine when the thread exits. 
This is also why I need to put dealloc into a block as well."

I reverted the mentioned change in the -dealloc and after quick testing, 
I can confirm that the exception goes away without the change in 
-dealloc. Scott, is it OK if we revert the change in -dealloc or it's 
needed to fix something else?

Thanks,
Dmitry

Anthony Petrov wrote:
> Hi Dmitry,
>
> The fix looks fine. I'm just wondering, do we really have to wrap the 
> -dealloc into perform...:wait:YES ? Is this a requirement from 
> JNFJObjectWrapper?
>
> -- 
> best regards,
> Anthony
>
> On 12/29/11 16:25, Dmitry Cherepanov wrote:
>> Hello,
>>
>> Please review a temporary fix for
>> http://java.net/jira/browse/MACOSX_PORT-686 at
>>
>> http://cr.openjdk.java.net/~dcherepanov/7124263/webrev.0/
>>
>> The fix ensures that -drawInCGLContext: and -dealloc is called on the
>> main thread (the fix is needed for the plugin and it's contributed by
>> Scott Kovatch). It's still unclear why -drawInCGLContext: isn't called
>> on the main thread and the change is suggested as a temporary 
>> workaround.
>>
>> Also, the patch includes some cleanup - it eliminates the code that
>> manages JNI global ref and starts using JNFJObjectWrapper.
>>
>> Thanks,
>> Dmitry
>>