The fix looks fine.
On 18.09.15 10:26, Semyon Sadetsky wrote:
On 9/18/2015 3:01 AM, Sergey Bylokhov wrote:
17.09.2015 8:33, Semyon Sadetsky wrote:
On 9/16/2015 5:54 PM, Sergey Bylokhov wrote:
16.09.2015 17:07, Semyon Sadetsky wrote:
Other scenarios do not cause conflicts.
Or we cannot reproduce it right now?
and for all these years...
But we got regression only now, because we are calling it from the
different places.
Direct call to DefWindowProc is much faster it's better to preserve
it.
Is the performance important in this case? This methods is used for
IME only? The unified implementation has a benefit that probability
to find a bug in the code is higher.
Those events come on each window activation regardless using IME. We
have a chance to get a performance regression in that case.
Do we know the root cause of the crash? Can we be sure that it cannot
be reproduced if our component is not a container? If it unclear then
it is better to use the slower but safer method.
So, slow but safe. I like it. See this:
http://cr.openjdk.java.net/~ssadetsky/8079595/webrev.01/
--Semyon
On 9/16/2015 4:52 PM, Sergey Bylokhov wrote:
Why we cannot always send a messages via SendMessage?
Can you add a comment to the changed lines to clarify the fix. It
is unclear for example why IsContainer() is used.
16.09.2015 15:36, Semyon Sadetsky wrote:
Hello,
Please review fix for JDK9:
bug: https://bugs.openjdk.java.net/browse/JDK-8079595
webrev: http://cr.openjdk.java.net/~ssadetsky/8079595/webrev.00/
Direct call to DefWindowProc() of the child window's owner causes
a conflict. Using SendMessage() eliminates it.
--Semyon
--
Best regards, Sergey.