review for 7090654: nightly failures after 7086585
tom.rodriguez at oracle.com
Wed Sep 14 13:36:21 PDT 2011
On Sep 14, 2011, at 1:31 PM, Vladimir Kozlov wrote:
> Looks good.
> We should also fix error message to print actual index and class name:
> fatal error: Invalid constant pool index %u for field name in class file %s
> But it could wait.
Yes, I noticed that too.
> Tom Rodriguez wrote:
>> 21 lines changed: 0 ins; 6 del; 15 mod; 2496 unchg
>> 7090654: nightly failures after 7086585
>> I had made late changes to accessing of fields in InstanceKlass and
>> didn't rerun my tests so I create a mismatch between the indexing.
>> Some code was using i * 7 and others was using i. I've corrected it
>> so it's using i consistently everything. Tested with sajdi.
>> I screwed up the new of fields when we recreate a .class file from our
>> internal representation. It should be unscaled. Tested with
>> java/lang/instrument regression tests. I had thought the tests I
>> originally ran exercised this path but I was wrong.
More information about the hotspot-compiler-dev