[7u6] Please review my fix for 7171163: [macosx] Shortcomings in the design of the secondary native event loop made JavaFX DnD deadlock (original) (raw)

steve.x.northover at oracle.com steve.x.northover at oracle.com
Thu May 24 08:26:34 PDT 2012

Thanks so much for finding and fixing this critical issue.

Steve

On 24/05/2012 10:33 AM, Alexander Zuev wrote:

Steve,

i guess you are right and i gladly borrow your version of comment to use. This and two other typos caught by Scott Palmer leads us to the fourth version of the webrev which can be found here: http://cr.openjdk.java.net/~kizune/7171163/webrev.03 With best regards, Alex On 5/24/12 17:48, steve.x.northover at oracle.com wrote: Thanks for commenting this for the next guy.

Nit pick: // Despite the naming this method on MacOS only making one event from // the native loop to be executed so this call is not blocking Is not great English. How about "Despite the naming of this method, on MacOS only one event is read and dispatched before this method returns. This call is non-blocking and does not wait for an event" Thanks, Steve On 24/05/2012 5:35 AM, Alexander Zuev wrote: Steve,

i've updated comments anywhere i can to reflect specific design of this feature on Mac. Also i have removed the unneeded call to the native function from the exit() method to save us some cycles on Java<->JNI call. Updated webrev can be found here: http://cr.openjdk.java.net/~kizune/7171163/webrev.01/ I also will send the formal request to approve the same change in jdk8 since we have to push change there before we can push it into the jdk7u6. With best regards, Alex On 5/23/12 22:26, steve.x.northover at oracle.com wrote: ToolkitThreadBlockedHandler seems to be shared code and it follows the enter / exit naming that is wrong. It seems that a lot of changes will be needed in shared code so I'm not sure the renaming is worth it.

You could add update the comments to indicate how this is used and save the renaming for another day. Steve On 23/05/2012 2:11 PM, Alexander Zuev wrote: Steve,

you are absolutely right - this native method is used only by our implementation of the DnD so i may just rename it and the only reason i have not done it is to minimize the scope of the change. So if you think renaming the method is required i will gladly do it. With best regards, Alex On 5/23/12 21:21, steve.x.northover at oracle.com wrote: The implementation matches how these methods are used by drag and drop but the naming makes no sense. Who else calls these methods? I'm assuming nobody (or that they are all called with the same implementation in mind).

startNativeNestedEventLoop is really "read and dispatch an event without blocking". Steve On 23/05/2012 12:51 PM, Alexander Zuev wrote: Hello,

please review my fix for the 7171163: [macosx] Shortcomings in the design of the secondary native event loop made JavaFX DnD deadlock Bug can be found at http://bugs.sun.com/bugdatabase/viewbug.do?bugid=7171163 Webrev can be found at: http://cr.openjdk.java.net/~kizune/7171163/webrev.00/ With best regards, Alexander Zuev

More information about the awt-dev mailing list