On 3/30/2011 8:48 AM, Tom Rodriguez wrote:
On Mar 30, 2011, at 7:38 AM, Daniel D. Daugherty wrote:
Added serviceability-dev@... since this affects SA code...
On 3/29/2011 11:58 AM, Tom Rodriguez wrote:
http://cr.openjdk.java.net/~never/7031614
Please update copyright years as needed.
I did.
Thanks!
I could be wrong here but some of what you've uncovered here
appears to be applicable to older SA (pre-static field move).
In particular the oop sizing stuff and static field visitation
during mirror processing.
It's not applicable to older ones. It used to get processed through the instance klass but when I moved them to Class they weren't being visited since I hadn't updated the SA.
Not even the oop size stuff? OK. Time for more coffee... :-)
Did you also get a chance to run the nsk.sajdi subsuite?
Yes. It was clean as well. None of these paths appear to be exercised by sajdi or I would have found it earlier.
Right after I sent the e-mail, I remembered that I didn't see
anything strange in nsk.sajdi during my weekend testing of the
latest bits here in Colorado. Nor was there anything strange
(beyond the usual stuff :-)) in RT_Baseline nightly over the
weekend.
This bug reminds that I need to check with Nicolay as to
whether the tmtools subsuite is ready for inclusion in
RT_Baseline nightly. I'll also have to add it to my mix here
in Colorado.
Thanks for being so diligent and for jumping on these new
failures so quickly!
Dan
Thumbs up.
Thanks!
tom
Dan
agent/src/share/classes/sun/jvm/hotspot/oops/ObjectHeap.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/oops/Instance.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/oops/InstanceKlass.java
line 478 - delete blank line?
agent/src/share/classes/sun/jvm/hotspot/oops/InstanceMirrorKlass.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/jdi/ClassObjectReferenceImpl.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/oops/Oop.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/oops/java_lang_Class.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/runtime/VM.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/tools/FinalizerInfo.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/utilities/HeapGXLWriter.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/utilities/HeapHprofBinWriter.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/utilities/ReversePtrsAnalysis.java
No comments.
agent/src/share/classes/sun/jvm/hotspot/utilities/soql/JSJavaFactoryImpl.java
No comments.
src/share/vm/oops/instanceMirrorKlass.hpp
No comments.
src/share/vm/runtime/vmStructs.cpp
No comments.
7031614: jmap -permstat fails with java.lang.InternalError in sun.jvm.hotspot.oops.OopField.getValue
Reviewed-by:
There were some places that were finding static fields by iterating
over the InstanceKlass object which I missed in my original changes.
I added an iterateStaticFields that hides the owner the static fields
and uses the OopVisitor.getObj to return the proper object. I
inspected all the iterateFields call sites to make sure I didn't miss
any. Additionally iteration of the mirror wasn't visiting the static
fields so some roots might be missed by the reverse pointers analysis.
There is heap iteration code in the SA and the InstanceMirrorKlass
wasn't plugged into the the oop sizing logic in the SA which caused
some searches to fail. I needed to add access to oop_size so I moved
the java_lang_Class style utilities into their own class and made them
like normal SA style classes. Tested with the full tmtools suite that
showed these issues plus some other which don't appear to have been
reported.
|
