Bug ID: JDK-5073483 JDWP: FATAL ERROR in native method swing+jvmstat(1.4.2 backend util.c line 1392)

Type: Bug
Component: core-svc
Sub-Component: debugger
Affected Version: 1.4.2

Priority: P4
Status: Closed
Resolution: Cannot Reproduce
OS: generic
CPU: generic

Submitted: 2004-07-13
Updated: 2004-11-11
Resolved: 2004-11-11

Customer reported the following error while monitoring a Swing application
with jvmstat while running under the Eclipse IDE debugger:

FATAL ERROR in native method: JDWP "util.c" (Feb 22 2004), line 1392: Unexpected error, error code = 21 (JVMDI_ERROR_INVALID_CLASS) 

Customer reports that the problem is difficult to reproduce.  The problem
occured with the following release:

java version "1.4.2_04"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_04-b05)
Java HotSpot(TM) Client VM (build 1.4.2_04-b05, mixed mode)

EVALUATION I've looked over the jdwp backend code for 1.4.2, and for this error to happen the jclass object would really need to have gone bad in-between calls to JVMDI. e.g., cycle over classes, one is bad. Since we held a reference (local or global), a GC shouldn't in theory mess us up. A retrace from a core file would be helpful, so I could see what the calls to isArrayClass() look like, that would help me narrow down when this is happening. If it happens with 1.5.0, I'd be very interested in looking at that. I have to assume that something is clobbering a jclass variable at this point, but that is a wild guess. ###@###.### 2004-07-13 If I get any more information on this I will investigate. ###@###.### 2004-07-28 The option -Xcheck:jni might help find any invalid use of JNI handles (jclass). Not much data to go on with this bug. ###@###.### 2004-11-10 23:06:50 GMT Closing out as not reproducable, sorry. Please re-file if this problem is seen again. ###@###.### 2004-11-11 21:43:07 GMT

10-11-2004

SUGGESTED FIX

29-07-2004

PUBLIC COMMENTS

29-07-2004