Javac Bug/Specification Question

Maurizio Cimadamore maurizio.cimadamore at oracle.com
Wed Apr 27 01:02:31 PDT 2011


Hi Richard,
this bug is a duplicate of 6650759 - good news is that it has been fixed 
in JDK 7; bad news is that it cannot be backported, as it involves some 
changes in the specification for generic method type-inference.

Thanks
Maurizio

On 07/04/11 00:22, Richard Warburton wrote:
> Hello,
>
> I hope this is an acceptable place to ask about this issue.  I would
> report a bug, except that I'm not entirely sure it is a bug, and I've
> seen similar issues discussed on here before.  Essentially I have a
> program that compiles fine in Eclipse, but doesn't using javac.  I
> suspect this is due to a generics/inference bug.  The problem is that
> when inferring a type of Foo<? super T>  that is passed a instance some
> concrete type javac infers the type Foo<Object>.  I have put a
> testcase online at:
>
> https://github.com/RichardWarburton/Compiler-Question
>
> This contains a 'build.sh' script that for me provides the following error:
>
> src/JavacFail.java:18:
> filter(java.util.List<java.lang.Integer>,Predicate<java.lang.Integer>)
> in JavacFail cannot be applied to
> (java.util.List<java.lang.Integer>,Predicate<java.lang.Object>)
>                  System.out.println(filter(asList(1, 2, 3), or(gt(2), lt(2))));
>                                     ^
> Note: src/JavacFail.java uses unchecked or unsafe operations.
> Note: Recompile with -Xlint:unchecked for details.
>
> I've also committed in the eclipse project files (in spite of usual
> practice) in an effort to make it easier to test that it works under
> ecj.
>
> I've tested under both javac 1.6.0_13 and javac 1.6.0_20, the version
> of eclipse that I'm using is Helios Service Release 1, build-id:
> 20100917-0705.
>
> If it helps in the detection of the bug it is possible to work around
> the issue by assigning the function to a seperate variable with the
> type, 'Predicate<Integer>', however I consider this a bit of an
> inferior solution.
>
> regards,
>
>    Richard




More information about the compiler-dev mailing list