RFR: 8142334: Improve lazy initialization of java.lang.invoke

[Claes Redestad]

Hi,

On 2015-11-12 19:10, Paul Sandoz wrote:
> Hi Peter,
>
> This stuff always gives me a headache :-)

Yes. :-)

>
> IIUC it’s all idempotent stuff, and the final field in NamedFunction should take care of certain things.

That's been my understanding, as well.

>
> Claes, was it intentional that you call function.resolve() after the array store? You might need to reverse that and place a Unsafe.storeFence between them if it is required that the published and visible function be resolved.

This however was unintentional, and I think you're right that this would 
ensure visibility:

         function.resolve();
         UNSAFE.storeFence();
         FUNCTIONS[idx] = function;

http://cr.openjdk.java.net/~redestad/8142334/webrev.04/

/Claes

>
> Paul.
>
>> >On 12 Nov 2015, at 17:26, Peter Levart<peter.levart at gmail.com>  wrote:
>> >
>> >Hi Claes,
>> >
>> >I have one concern...
>> >
>> >645     private static NamedFunction getConstantFunction(int idx) {
>> >646         NamedFunction function = FUNCTIONS[idx];
>> >647         if (function != null) {
>> >648             return function;
>> >649         }
>> >650         return setCachedFunction(idx, makeConstantFunction(idx));
>> >651     }
>> >652
>> >653     private static synchronized NamedFunction setCachedFunction(int idx, final NamedFunction function) {
>> >654         // Simulate a CAS, to avoid racy duplication of results.
>> >655         NamedFunction prev = FUNCTIONS[idx];
>> >656         if (prev != null) {
>> >657             return prev;
>> >658         }
>> >659         FUNCTIONS[idx] = function;
>> >660         function.resolve();
>> >661         return function;
>> >662     }
>> >
>> >
>> >Above is a classical double-checked locking idiom, but it is not using volatile variable to publish the NamedFunction instance. I wonder if this is safe. Even if the FUNCTIONS[idx] slot was a volatile variable,  you would publish new instance before resolving it. Is it OK to publish unresolved NamedFunction(s)? There is a NamedFunction.resolvedHandle() instance method that makes sure NamedFunction is resolved before returning a MethodHandle, but there are also usages of dereferencing NamedFunction.resolvedHandle field directly in code. Are you sure that such unresolved or almost resolved instance of NamedFunction is never used in such places where NamedFunction.resolvedHandle field is dereferenced directly?
>> >
>> >In original code those NamedFunctions were resolved in static initializer so they were published properly.
>> >
>> >Regards, Peter