On 11/13/2012 4:47 PM, Anthony Petrov wrote:
Hi Alexander,
I'm wondering why does the bug synopsis mention a concrete platform
(Linux), but the fix is in shared code? Why wasn't this a problem for
e.g. Windows or Mac previously? And will anything be affected by this
change on these other platforms?
The AWT-Shutdown thread starts from WToolkit constructor (see the
fix for the issue 4701990). And the toolkit is created with the
necessary permissions in the Toolkit.getDefaultToolkit() method.
The AWT-Shutdown thread started from the native awt.m library on
MacOSX which is load with the necessary permissions.
The AWT-Shutdown thread starts on Linux from the main thread when
the system needed it and so it does not have necessary permissions.
The AWTAutoShutdown.notifyToolkitThreadBusy() method is public so
it can be invoked from any place on any OS and leads to the security
exception if a SecurityManager is set.
The current fix does not affect Windows and MacOSX platforms.
Thanks,
Alexandr.
--
best regards,
Anthony
On 11/13/2012 1:12 PM, Alexander Scherbatiy wrote:
On 11/12/2012 9:52 PM, Sergey Bylokhov wrote:
Hi, Alexander.
Could you review the updated fix:
http://cr.openjdk.java.net/~alexsch/8002191/webrev.01/
Does setDaemon also requires additional permissions?
Yes. It requires the modifyThread permission.
Also note, all this file has 80 lines alignment.
The long line is formatted.
Thanks,
Alexandr.
12.11.2012 18:32, Alexander Scherbatiy wrote:
bug: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8002191
webrev: http://cr.openjdk.java.net/~alexsch/8002191/webrev.00
The AWT-Shutdown thread starts from the main thread on Linux and so
it does not have modifyThreadGroup permissions.
The fix runs the AWT-Shutdown thread from the doPrivileged block.
The 'thread.setDaemon(false);' line is not formatted because hg
diff (and so webrev) does not treat new spaces as a change.
Thanks,
Alexandr.
