compiler accepts incompatible method reference for Sam conversion

[John Coomes]

maurizio cimadamore (maurizio.cimadamore at oracle.com) wrote:
> On 16/07/2011 08:59, Ali Ebrahimi wrote:
> > Hi,
> > for me this still crash with full lambda repo build.
> > The main reason for this, i think, is that hotspot sub-repo of lambda 
> > repository is not sync with jdkb147(currently, latest tag is jdkb143).
> Yeah - I noticed the weird tag in hotspot - I merged it against the JDK 
> 8 repo, which seems to have the same problem (all subrepos have b147 tag 
> but hotspot, which has b143).

The hotspot in jdk8 is missing some fixes that went in late to jdk7 as
well as the latest build tags.  This is due to the way changes were
restricted/managed at the end of jdk7.  The missing fixes should be
pushed up to jdk8/jdk8/hotspot next week.  The fate of the tags is
less clear, since the hotspot repo in jdk7 diverged from jdk8.  I'll
see what can be done about the tags.

-John

> I will investigate about this.
> 
> Maurizio
> >
> > Best Regards,
> > Ali Ebrahimi
> >
> > On Wed, Jul 13, 2011 at 1:22 PM, Maurizio Cimadamore 
> > <maurizio.cimadamore at oracle.com 
> > <mailto:maurizio.cimadamore at oracle.com>> wrote:
> >
> >     Ali,
> >     this works for me (using b147 - or the VM bundled in the lambda
> >     repo). There have been tons of fixes in the MH area lately, so
> >     make sure you grab a more recent JDK 7 snapshot (I used to see
> >     those problems too before the latest lambda merge, but now ost of
> >     them went away).
> >
> >     Note also that, in cases in which 292 code is still unstable, you
> >     can switch back to the old generation logic using the -XDuseProxy
> >     flag, which should remove all 292-related problems (as, by doing
> >     so, java.lang.reflect.Proxy will be used instead).
> >
> >     I will be pushing a fix for the other issue (incompatible method
> >     ref erroneously accepted by javac) shortly.
> >
> >     Thanks
> >     Maurizio
> >
> >
> >     On 13/07/11 10:36, Ali Ebrahimi wrote:
> >
> >         Hi Remi,
> >
> >         On 7/13/11, Rémi Forax<forax at univ-mlv.fr
> >         <mailto:forax at univ-mlv.fr>>  wrote:
> >
> >             On 07/13/2011 06:01 AM, Ali Ebrahimi wrote:
> >
> >             Hi Ali,
> >
> >             ...
> >
> >                 MethodHandle mh =
> >                 MethodHandles.lookup().findStatic(MethodHandleTest.class,
> >                 "sTest",
> >                 MethodType.methodType(void.class,int.class,int.class));
> >                 MethodHandle smh = mh.asSpreader(Object[].class,
> >                 mh.type().parameterCount());
> >
> >                 smh.invoke(new Object[]{i,i});<===============crash
> >
> >                 main cause: mh.asSpreader
> >
> >                 I don't known this crash appears in jdkb147.
> >
> >             What you want here is to use asCollector, not asSpreader.
> >             asSpreader requires that the method as already an object array
> >             and extract the value from it.
> >
> >         May be i'm wrong. but this code picked from
> >         MethodHandleProxies class
> >
> >             public static
> >         <T>  T asInterfaceInstance(final Class<T>  intfc, final
> >         MethodHandle target) {
> >                 final Method sm = getSingleMethod(intfc);
> >                 if (sm == null)
> >                     throw new IllegalArgumentException("not a
> >         single-method
> >         interface: "+intfc.getName());
> >                 MethodType smMT =
> >         MethodType.methodType(sm.getReturnType(),
> >         sm.getParameterTypes());
> >                 MethodHandle checkTarget = target.asType(smMT);  //
> >         make throw WMT
> >                 checkTarget =
> >         checkTarget.asType(checkTarget.type().changeReturnType(Object.class));
> >                 final MethodHandle vaTarget =
> >         checkTarget.asSpreader(Object[].class, smMT.parameterCount());
> >                 return intfc.cast(Proxy.newProxyInstance(
> >                         intfc.getClassLoader(),
> >                         new Class[]{ intfc, WrapperInstance.class },
> >                         new InvocationHandler() {
> >                             private Object getArg(String name) {
> >                                 if ((Object)name ==
> >         "getWrapperInstanceTarget")  return target;
> >                                 if ((Object)name ==
> >         "getWrapperInstanceType")
> >           return intfc;
> >                                 throw new AssertionError();
> >                             }
> >                             public Object invoke(Object proxy, Method
> >         method,
> >         Object[] args) throws Throwable {
> >                                 if (method.getDeclaringClass() ==
> >         WrapperInstance.class)
> >                                     return getArg(method.getName());
> >                                 if (method.equals(sm))
> >                                     return vaTarget.invokeExact(args);
> >                                 if (isObjectMethod(method))
> >                                     return callObjectMethod(this,
> >         method, args);
> >                                 throw new InternalError();
> >                             }
> >                         }));
> >             }
> >
> >         and this results in this crash:
> >         class MethodHandleTest{
> >         ...
> >         public static void sTest(int x,int y){
> >              //   System.out.println("stest");
> >             }
> >
> >           static interface Sam{
> >                 void m(int x,int y);
> >             }
> >         ..
> >         Sam sam = MethodHandleTest#sTest;
> >                 sam.m(1,1);<==========
> >
> >             In your example, the call to mh.asSpreader should return a
> >             runtime
> >             exception,
> >             I will investigate why it's not the case when I will be at
> >             my office :)
> >
> >                 Best Regards,
> >                 Ali Ebrahimi
> >
> >             regards,
> >             Rémi
> >
> >
> >             _______________________________________________
> >             mlvm-dev mailing list
> >             mlvm-dev at openjdk.java.net <mailto:mlvm-dev at openjdk.java.net>
> >             http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
> >
> >         Best Regards,
> >         Ali Ebrahimi
> >
> >
> >
> 
>