Bugzilla – Full Text Bug Listing
| Summary: | object reference graph incorrect | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Tools] MAT | Reporter: | Randall Theobald <rstheo> | ||||
| Component: | Core | Assignee: | Project Inbox <mat.core-inbox> | ||||
| Status: | RESOLVED NOT_ECLIPSE | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | andrew_johnson | ||||
| Version: | 0.8 | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | PC | ||||||
| OS: | Windows XP | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
I can reproduce the problem with objects 0xdc561b8 0xe06adb8 0xe0653f0 0xdb851c8 The dump is partially corrupt as in the error log there are 20 messages of these types Invalid object field reference 0xdb89300 of type com.ibm.dtfj.image.CorruptDataException: Unknown class ID 230216736 found at field java/util/WeakHashMap$Entry.value of type Ljava/lang/Object; in object of type java/util/WeakHashMap$Entry at address 0xdb851c8 while finding outbound references Invalid array element reference 0xdb894b0 of type ? found at index 0 in array of type [Ljava/lang/Object; length 5 at address 0xdb853b0 while finding outbound references and this final message, so there have been hundreds of thousands of these errors. 297,876 repeated messages have been suppressed What VM options were used to generate the dump? My eclipsi.ini had this line: -Xdump:system:events=user+systhrow,filter=*Memory*,request=exclusive+prepwalk+compact but it was our Performance Inspector JProf JVMTI agent that triggered the system dump. How can I find those error messages? Can we put some sort of warning up in MAT to notify the user of such a thing? My jextract log was clean. Window->Error Log will show you the messages from the parser. The dump was corrupt, so we can't do anything. |
Created attachment 144269 [details] problem system dump I think I have come across what could be a major bug in MAT or the IBM adapter. In this particular dump I have (IBM system dump), I keep finding references that should be impossible. For example, in a WeakHashMap$Entry.next field, the object referred to was NOT a WeakHashMap$Entry object. I came across a few other instances where it made me question the integrity of the data being shown to me. Open the attached dump in MAT. Go to the class XSDParser, static field, userDataMap. Browse the WeakHashMap entries and see that there are some .next fields are not pointing to WeakHashMap$Entry objects. (e.g. 0xdc561b8, 0xe06adb8)