EVALUATION

Need more information. Without a test case or access to customer machines, or at
least more detailed information, impossible to make a diagnosis.

Try running with -Xint. If that works then the problem is in the generated code
for some Java method and we can narrow it down. If this doesn't help, run with
-Xcheck:jni. If errors are reported (unclear based on the bug report whether
there is native code in the customer's application) then these are the cause of
the problem and are in customer code.

###@###.### 2003-06-05

the dump with -Xcheck:jni failed on function: checkArray
when it is calling Java_java_net_SocketInputStream_socketRead0:

0150a07c 0eeff72c 00000008 jvm!checkArray+0x1f
0150a07c 0eeff72c 00000000 jvm!checked_jni_SetByteArrayRegion+0x27
WARNING: Stack unwind information not available. Following frames may be wrong.
0150a07c 0eeff734 00000f9c net!Java_java_net_SocketInputStream_socketRead0+0x118
0000ea60 00000005 00000000 0x1727624
....

The behavior is strange, see the attached file eval.txt. We have reason to question hardware problem here.


###@###.### 2003-06-26

The problem is in MS code, on the failure thread stack, the arguments from Java are two parameters: buffer and len, we allocate the buffer on stack, then pass them into Windows API. Just before calling into Win32 API, the first parameter, the buffer address on stack is NULL, but other parameters are correct: the len of the buffer and the return address in calling function. See attachment for the stack analysis.

###@###.### 2003-09-25