Solaris 8/SPARC JDK 1.2.2_15(java_g) crashes in GC.


---------------
Callback exceeded maximum time or time to stop
Card 80354 should not be empty: obj = fb73c428 (class java/lang/ref/Finalizer),
refPtr = fb73c438
  *** Garbage Collection in process, a thread
  *** dump is not possible.

*** panic: "../../../../src/share/javavm/runtime/gc/gen/cardtable.c",
line 642:
verification failure:
         expression: FALSE

SIGABRT 6 abort (generated by abort(3) routine)
     si_signo [6]: ABRT
     si_errno [0]:
     si_code [-1]: SI_LWP [pid: 18954, uid: 1010]
         stackpointer=FEAF0A68
  *** Garbage Collection in process, a thread
  *** dump is not possible.
--------

This phenomenon occurs within 1 hour after production system starts.
I had reproduced this phenomenon by same program as BugId 4846970.
However, reproducing by this program needs five to ten days.

And this phenomenon occurs on both standard libthread and alternate
libthread.


###@###.### 2003-06-25

I guess that this phenomenon has influence on memory size and thread count.
On production system, the rate of crash changes by change of max heap size
(Although the maximum specified, the minimum does not specify)

  -mx128m : 4 times/3 months
  -mx1024m : twice a year

On sample program, the rate of crash changes by change of number of threads.

  4 threads : two of 1000 times. (0.2%)
  8 threads : six of 78 times. (7.7%)


###@###.### 2003-08-13

Unfortunately, this problem reproduced on
Solaris_JDK_1.2.2_16+4846970+4887114+4890910_2003.07.17.

 12 threads : 2 of 150 times.
  8 threads : 8 of 150 times.

So, can I have you send me the patch list of your environment?