There seems to be a regression 1.3.1_09 and in 1.3.1_10. 
A Thread.interrupt() call no longer executes after first try.

Using 1.3.1_08 or version earlier the expected behavior is seem:

greatpoint> /net/koori.sfbay/a/v13/jdk/1.3.1_08/fcs/binaries/solaris-sparc/bin/java test.InterruptThread
 +++ Poller.run
 @@@ it is running
 +++ poller interrupted!
 @@@ it is running
 @@@ it is running
 +++ poller interrupted!
 @@@ it is running
 +++ poller interrupted!
 @@@ it is running
 +++ poller interrupted!

But running the same testcase with _09 and _10 the Thread.interrupt is passed
over:

greatpoint> /home/atongsch/JDK/1.3.1_09/j2sdk1_3_1_09/bin/java test.InterruptThread
 +++ Poller.run
 @@@ it is running
 +++ poller interrupted!
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
^Cgreatpoint> /net/koori.sfbay/p/v10/jdk/1.3.1_10/fcs/b02/binaries/solaris-sparc/bin/java test.InterruptThread
 +++ Poller.run
 @@@ it is running
 +++ poller interrupted!
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running
 @@@ it is running


The testcase is located in the attachment

N/A

Since this problem is reported as a regression between two update releases, 
I am routing the bug to the escallation team.

--  xxxxx@xxxxx  2003-11-17

I believe this regression is introduced by the asynchronous signal handling fix
in 1.3.1_09 in hotspot runtime system. So I change the category and subcategory to hotspot and runtime.


 xxxxx@xxxxx  2003-11-17

The previous backport fix of 4460034 did not set _interrupt_in_progress to false in Thread_Interrupt_Callback.

 xxxxx@xxxxx  2003-11-24

This thread_interrupt_callback doesn't seem to be a problem 
in a windows environment.  It's only in solaris environment 
where this problem exist.