Java Solaris Communities Sun Store Join SDN My Profile Why Join?
 
Bug Database
Bug Detail
Quick Lists
Top 25 Bugs
Top 25 RFE's
Recently Closed Bugs
Printable Page Printable Page


Bug Database
Bug ID: 4809618
Votes 0
Synopsis Lin64: ThreadInterruptedAndNotified test causes segmentation fault
Category hotspot:runtime_system
Reported Against mantis-beta
Release Fixed
State 11-Closed, Not Reproducible, bug
Priority: 2-High
Related Bugs 4701394 , 6733855 , 4916096
Submit Date 29-JAN-2003
Description
pwd-------------------------------------------------------
test:		runtime/threads/ThreadInterruptedAndNotified
testbase:	testbase_vm
JDK:		mantis b14
System:		RH7.2 ia64
VM:		Server
Mode:		Xmixed, Xcomp
-------------------------------------------------------
This test FAILED with mantis b11 on lin64
This test PASSED with mantis b14 on solaris 5.8 (x86), RH8.0(32bit), winXP64

The point of this test is that if two (or more) threads are
waiting on the same  customer  and one of them is simultaneously notified
and interrupted then the notification should be passed to one of
the other waiting threads.

Steps to reproduce 
================ 
1. cd /net/sqesvr.sfbay/export/vsn/GammaBase/Bugs/{BugID} 
2. sh rerun.sh 
Work Around
N/A
Evaluation
This test segv'ed on a dual processor itanium 1 using the kernel with the
MP problem.  Using the latest HP kernel caused it to print

Interrupt did not redirect notification.

which is still wrong.  The latter also happens on both single and dual processor
itanium 2's.

  xxxxx@xxxxx   2003-02-03

This test fails with b11, contrary to the description.

  xxxxx@xxxxx   2003-02-03


THe test also fails with b11.  It's a typo to the description.
I update it now.  thks,


  xxxxx@xxxxx   2003-02-03

With the patched MP kernel, the test now always produces
 
 	Interrupt did not redirect notification.
 	
Looking at ThreadInterruptedAndNotified.java (attached), this is a perfectly
legitimate result.  I.e., it's certainly possible for the interrupt
to be delivered an instance of the Notified class _after_ a notify.
In that case, the Notifier won't see the interrupt.

I tested this by commenting out the System.exit() after the
println that produces the above msg.  The test ran to completion and
printed PASS.

Hung Nguyen, the bug submitter, agrees, so I'm marking this bug NR.
I understand we're in the process of upgrading test machine kernels.
Without the kernel fix, this test will still fail.

  xxxxx@xxxxx   2003-02-05
Comments
  
  Include a link with my name & email   


PLEASE NOTE: JDK6 is formerly known as Project Mustang