RFR 8025937: assert(existing_f1 == NULL || existing_f1 == f1) failed: illegal field,change

[Coleen Phillimore]

On 11/13/2013 2:35 PM, John Rose wrote:
> That looks good.  The linear search is a fine choice for this corner case.

Yeah, there are few and not worth adding a ton of code and/or allocating 
another cp_map table for.
>
> BTW, I suggest adding an assert to cp_cache_delta() that _first_iteration_cp_cache_limit >= 0.

Thanks, that is a good suggestion.  I will add it.
>
> (Nit:  The "!!!" in the comment is a trifle loud.)

Ok. I'll take out two of the !..  :)

Thanks!
Coleen
>
> You can count me as reviewer in any case.
>
> — John
>
> On Nov 13, 2013, at 9:04 AM, Coleen Phillimore <coleen.phillimore at oracle.com> wrote:
>
>> I added code to so we don't create duplicate invokespecial/InterfaceMethodref cpCache entries (suggested by John R), renamed _cp_cache_index_limit to _first_iteration_cp_cache_limit which is a bit clearer, and fixed formatting in test.   Also added duplicate invokespecial/InterfaceMethodref case to test.   Reran jck, java/lang/invoke and defmeth tests.
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/8025937_2/
>>
>> Thanks,
>> Coleen
>>
>> On 11/10/2013 11:24 PM, Coleen Phillmore wrote:
>>> Summary: Create extra constant pool cache entries for invokespecial/InterfaceMethodref to hold the alternate resolution.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8025937/
>>> bug link https://bugs.openjdk.java.net/browse/JDK-8025937
>>>
>>> Tested with testcase in time api, java/lang/invoke jtreg tests, NSK mlvm, quick and defmeth tests, jck lang/vm tests, and testcase in changeset.
>>>
>>> Thanks,
>>> Coleen