<AWT Dev> <AWT dev>[9] Review request for JDK-8165555: [macosx] VM crashes on second attempt to execute JCK interactive tests that use Robot (single JVM, agent)

[Sergey Bylokhov]

Hi, Manajit. 
It looksbetter, bet it can be improved a little bit. 
- you should not skip exceptions in the run(); 
- It will be good to select the frame by mouse click before pressing the key, otherwise you can run somthing else. 
- I suggest to iterate the code in main a few times in the loop(I guess 10 iterations is ok) 
-The "frame" can be a local var inside the createRobot(), also the method can be renamed. 

----- manajit.halder at oracle.com wrote: 
> 
> Hi Sergey, 
> 

> Thanks for the review. Changes in CRobot.m are removed and added a jtreg test case. 
> Please review the webrev. 
> 

> http://cr.openjdk.java.net/~mhalder/8165555/webrev.03/ 
> 

> Thanks, 
> Manajit 
> 




> On 27-Sep-2016, at 4:15 pm, Sergey Bylokhov < Sergey.Bylokhov at oracle.com > wrote: 

> On 27.09.16 11:51, Manajit Halder wrote: 


Hi Sergey, 

Thanks for the comment. “self.javaToMacKeyMap is retaining the reference. 
Please review the modified code: 

http://cr.openjdk.java.net/~mhalder/8165555/webrev.02/ 

The changes in CRobot.m is unnecessary? I also suggest to write the test. Recently I have run all jck tests on my osx system and no of them were crashed, so it seems that the sequence of tests in the bug report are not stable and the new reg test will be welcome to catch such bugs always. 





On 23-Sep-2016, at 10:24 pm, Sergey Bylokhov 
< Sergey.Bylokhov at oracle.com < mailto:Sergey.Bylokhov at oracle.com >> wrote: 

On 21.09.16 16:01, Manajit Halder wrote: 


Hi Sergey, 

Thanks for the comment. 

Access to "instance" is not broken. The problem is with the dictionary 
variable "javaToMacKeyMap" within the "instance" reference. 
The dictionary is not getting initialised for the second time when 
singleton method is called for the second time. The dictionary is 
getting initialised during the first time singleton method is called and 
hence crash was observed. Tried adding self.javaToMacKeyMap 
but doesn’t solve the problem. 

I still suggest you to check "self.javaToMacKeyMap" at line 48 this 
should call generated setter which will retain the reference. 




I propose two solutions to this problem: 

Solution 1: 
Reinitialise the dictionary every-time the singleton method is called. 
Webrev: 
http://cr.openjdk.java.net/~mhalder/8165555/webrev.01/ 

Drawback: 
dictionary is initialised multiple times (every time singleton method is 
called). 

Solution 2: 
Make the dictionary static. 
Drawback: 
Still dictionary need to be initialised multiple times. 
No singleton method, just two static methods, one method to initialise 
the dictionary and other to get the key form the dictionary. 
Webrev: 
Not prepared. Will be prepared if second solution accepted. 

Test case: 
There are no test cases. Problem can be reproduced by 
executing following JCK test: 
java -jar JCK-runtime-9/lib/jtjck.jar -mode:single -k:interactive 
"api/java_awt/interactive/event/EventTests.html#EventTest0019 
api/java_awt/interactive/event/EventTests.html#EventTest0013” 

Thanks, 
Manajit 



On 20-Sep-2016, at 4:42 am, Sergey Bylokhov 
< Sergey.Bylokhov at oracle.com 
< mailto:Sergey.Bylokhov at oracle.com > < mailto:Sergey.Bylokhov at oracle.com >> 
wrote: 

Hi, Manajit. 
It seems that after the fix "(CRobotKeyCodeMapping *) sharedInstance" 
returns the new object per invocation, so it is not really 
sharedInstance. I am not sure I understand what is wrong in the 
current code, from the my point of view this is a correct singleton. 
It it true that the problem is in access to broken "instance" and not 
to "javaToMacKeyMap" inside the "instance"? If not then 
"javaToMacKeyMap" should be changed to "self.javaToMacKeyMap". 
Do you have a test case to reproduce the bug? 

On 19.09.16 15:26, Manajit Halder wrote: 


Hi All, 

Kindly review the fix for JDK9. 

Bug: 
https://bugs.openjdk.java.net/browse/JDK-8165555 

Webrev: 
http://cr.openjdk.java.net/~mhalder/8165555/webrev.00/ 

Issue: 
[macosx] VM crashes on second attempt to execute JCK interactive tests 
that use Robot (single JVM, agent) 

Cause: 
While executing the JCK test for the second time the robot was getting 
initialised once again and old instance of CRobotKeyCodeMapping was not 
available. 
Crash was observed while trying to access invalid instance of 
CRobotKeyCodeMapping. 

Fix: 
A new instance of CRobotKeyCodeMapping is created when robot is 
initialised. 

Regards, 
Manajit 


-- 
Best regards, Sergey. 



-- 
Best regards, Sergey. 



-- 
Best regards, Sergey. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/awt-dev/attachments/20161004/5b8ad558/attachment.html>