Name: elR10090 Date: 12/17/2001

-------------------------------------
  Test : nsk/jdwp/Event/VM_DEATH/vmdeath001 nsk/jdi/EventQueue/remove_l/remove_l003
        nsk/jdi/VirtualMachine/exit/exit001

  TestBase : testbase_nsk
  VM : server
  Mode : int
  Platform : Itanium
  OS : WINXP
  
 ----------------------------------------
Ivan Popov <###@###.###>

The following JDWP test

    nsk/jdwp/Event/VM_DEATH/vmdeath001

failed against JDK 1.4.0-rc1-b89 on Windows.

This test uses raw socket channel for JDWP connection
and verifies if JDWP event VM_DEATH is received when
debuggee VM exits.

The test launches debuggee VM, waits for VM_START event
is received and resumes debuggee VM. After resuming
the debuggee VM just exits and the test waits for
VM_DEATH event packet.

However, if debuggee VM has time to finish before the test
tries to read VM_DEATH event packet from JDWP channel, a
SocketException is thrown with the diagnostics:

  java.net.SocketException:
      Connection reset by peer: JVM_recv in socket input stream read

But if the test has time to read VM_DEATH event packet before
debuggee VM actually exited, no exception is thrown.

I slightly modified the test by including sleep for 10 seconds
before reading VM_DEATH event packet to make this failure
more stable.

This failure occurs only on Windows. On the all ather platforms
this test pass, even with the delay before reading event.

Here is citation from the test log (on Windows):
================================================================
Waiting for VM_INIT event
  ... VM_INIT event received
Querying for IDSizes
  ... size of VM-dependent types adjusted
Resumindg debuggee
  ... debuggee resumed
Sleeping for 10 seconds
debugee.stderr> Debugee exited
  ... continued after sleep

>>> Testing JDWP event

Waiting for event packet
# ERROR: Unable to read tested event packet:
# ERROR: java.net.SocketException: Connection reset by peer:
                JVM_recv in socket input stream read
================================================================

This failure affects many JDWP tests of testbase_nsk because
they use VM_DEATH event to be notified about debuggee VM shutdown.

These tests will appear in the next release of testbase_nsk
located here:

    /net/sqesvr.sfbay/export/vsn/VM/testbase/testbase_nsk.v14

To reproduce the bug run 'doit.sh' script located here:

    /net/sqesvr.sfbay/export/vsn/GammaBase/Bugs/<this bug number>

Use:
    sh doit.sh $JAVA_HOME
Where:
    $JAVA_HOME: points to JDK 1.4-b89 for Windows

======================================================================

As noted here and elsewhere the issue is that the transport isn't
closed gracefully when the VM terminates gracefully. This issue is
understood and is being fixed in tiger as 4816980.

###@###.### 2003-09-04