RFR(S): 8157225: adopt method handle for array length getter in BeanLinker

[Michael Haupt]

Hi Attila,

thanks for your comments.

> Am 18.05.2016 um 16:11 schrieb Attila Szegedi <szegedia at gmail.com>:
> There’s a bit of an issue with this. Namely, the old GET_ARRAY_LENGTH being the unreflection of j.l.reflect.Array.getLength() worked for any array type, so it was always accompanied with ValidationType.IS_ARRAY. I believe this is wrong now and can cause ClassCastException at runtime (I say “believe” as I can’t currently build jdk 9 so can’t confirm). This could be caught by a test that handles two arrays of different types in the same call site, e.g. with Nashorn:
> 
>  var intArray = Java.type(“int[]”)
>  var doubleArray = Java.type(“double[]”)
>  var arrs = [new intArray(0), new doubleArray(0)]
>  for (var i in arrs) arrs[i].length

Confirmed:

$ jjs
jjs> var intArray = Java.type("int[]")
jjs> var doubleArray = Java.type("double[]")
jjs> var arrs = [new intArray(0), new doubleArray(0)]
jjs> for (var i in arrs) arrs[i].length
java.lang.ClassCastException: Cannot cast [D to [I

> Since the MethodHandles.arrayLength is typed to the concrete array class, it should be used with ValidationType.EXACT_CLASS.

Also confirmed:

$ jjs
jjs> var intArray = Java.type("int[]")
jjs> var doubleArray = Java.type("double[]")
jjs> var arrs = [new intArray(0), new doubleArray(0)]
jjs> for (var i in arrs) arrs[i].length
0

> This will obviously cause call sites taking length of different kinds of arrays more polymorphic… FWIW, the IS_ARRAY validation type is probably obsolete now as it was only used for array length getters anyway.

Yes; there are two uses of IS_ARRAY that can be replaced with EXACT_CLASS ...

> The polymorphism could be somewhat alleviated by having stable validation for all arrays of reference types. Maybe have one MethodHandles.arrayLength(Object[].class) handle cached to use with all arrays of reference types, and use a Validator(Object[].class, ValidationType.INSTANCE_OF) with it. Then you’d have stable linkage that can take the length of any reference-typed array, but you’d still trigger relinking for primitive-typed arrays.

... but shouldn't, right.

I've filed a bug to adopt the quick fix with EXACT_CLASS [1], and another to deal with the relinking problem [2]. The solution you suggest should be possible; maybe the IS_ARRAY abstraction can be reused for this in a way that avoids relinking for primitive arrays by having a PIC-like structure.

Best,

Michael

[1] https://bugs.openjdk.java.net/browse/JDK-8157250
[2] https://bugs.openjdk.java.net/browse/JDK-8157251


-- 

 <http://www.oracle.com/>
Dr. Michael Haupt | Principal Member of Technical Staff
Phone: +49 331 200 7277 | Fax: +49 331 200 7561
Oracle Java Platform Group | LangTools Team | Nashorn
Oracle Deutschland B.V. & Co. KG | Schiffbauergasse 14 | 14467 Potsdam, Germany

ORACLE Deutschland B.V. & Co. KG | Hauptverwaltung: Riesstraße 25, D-80992 München
Registergericht: Amtsgericht München, HRA 95603

Komplementärin: ORACLE Deutschland Verwaltung B.V. | Hertogswetering 163/167, 3543 AS Utrecht, Niederlande
Handelsregister der Handelskammer Midden-Nederland, Nr. 30143697
Geschäftsführer: Alexander van der Ven, Jan Schultheiss, Val Maher
 <http://www.oracle.com/commitment>	Oracle is committed to developing practices and products that help protect the environment