The dead-lock detection feature should be added to JVM_DumpAllStacks().

When there is dead-lock in Java, the status of dead-lock
is reported(output) in Full thread Dump with SIGQUIT.

This process seems to be implemented at the following (*)
extracted from ./share/vm/runtime/os.cpp.
(This example is based on 1.4.2_04)

----->
......
    switch (sig) {
      case SIGBREAK: {
        // Print stack traces
        VM_PrintThreads op;
        VMThread::execute(&op);
        VM_FindDeadlocks op1; (*)
        VMThread::execute(&op1); (*)
        if (PrintClassHistogram) {
          VM_GC_HeapInspection op1;
          VMThread::execute(&op1);
        }
        jvmpi::post_dump_event();
        break;
      }
....
<------

On the other hand, there seems the similar feature to output
Full Thread Dump in JavaVM function, JVM_DumpAllStacks().
However, this function does not seem to work, which is,
not detect dead-lock status.
(The following src is extracted from . src/share/vm/prims/jvm.cpp.)

--->
....
JVM_ENTRY(void, JVM_DumpAllStacks(JNIEnv* env, jclass))
  JVMWrapper("JVM_DumpAllStacks");
  VM_PrintThreads op;
  VMThread::execute(&op);
  jvmpi::post_dump_event();
JVM_END
...
<-----

JVM_DumpAllStacks() seems called by Plug-in.
If true, when a dead lock occurs during plug-in,
dead-lock detection feature will not work correctly.

SUGGESTION:
  How about to apply the following modification to jvm.cpp
  as follows ?

----->
*** jvm.cpp.orig Thu Jun 17 19:36:45 2004
--- jvm.cpp Thu Jun 17 19:38:49 2004
***************
*** 2144,2149 ****
--- 2144,2151 ----
    JVMWrapper("JVM_DumpAllStacks");
    VM_PrintThreads op;
    VMThread::execute(&op);
+ VM_FindDeadlocks op1;
+ VMThread::execute(&op1);
    jvmpi::post_dump_event();
  JVM_END
<----

######################################################################