RFR (S) 8197780: Null pointer dereference in Klass::is_instance_klass of klass.hpp:532

[coleen.phillimore at oracle.com]

On 2/22/18 11:42 AM, Lois Foltan wrote:
> On 2/22/2018 10:46 AM, coleen.phillimore at oracle.com wrote:
>
>> Summary: Make generate_class_cast_message able to deal with NULL 
>> klass from the constant pool.
>>
>> Also moved the function to get the class loader/module name to Klass 
>> where it belongs and removed an unused function in the constant pool 
>> that looks like the one I ended up using.
>>
>> Tested with tier1 and which includes the test added for the original 
>> change and compiler test that has null klass in the constant pool 
>> entry: compiler/escapeAnalysis/cr6716441/Tester.java
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/8197780.01/webrev
>> bug link https://bugs.openjdk.java.net/browse/JDK-8197780
>
> Looks good.  I like the move of class_loader_and_module_name to 
> Klass.  Please consider
>
> - adding the same comment about the caller (or one of its callers) 
> must use a ResourceMark
>   ahead of the SharedRuntime::generate_class_cast_message(Klass*, 
> Klass*, Symbol*) method.
Okay, now all three have that comment!
> - adding an assert to make sure if target_klass == NULL, then 
> target_klass_name != NULL.

Okay, I'll make that change too, and rerun small test to make sure this 
is right:

+  assert(target_klass != NULL || target_klass_name != NULL, "one must 
be provided");


Thanks!
Coleen
>
> I don't need to see another webrev.
>
> Thanks,
> Lois
>
>>
>> Thanks,
>> Coleen
>