Misleading error message when using diamond operator with private constructor

[Maurizio Cimadamore]

On 21/08/13 19:32, Eddie Aftandilian wrote:
> Hi compiler-dev,
>
> I work with Jeremy on javac at Google.  A user pointed out to us that
> compiling code that uses the diamond operator and attempts to access a 
> private constructor fails with a misleading error message:
>
> Test.java:3: error: cannot infer type arguments for MyList<>
>     MyList<Object> foo = new MyList<>();
>                          ^
> 1 error
>
> The error is not that the types cannot be inferred, but that the 
> constructor for MyList is private.  If you explicitly specify the type 
> parameter in the call to the MyList constructor, you get the correct 
> error message telling you that the constructor is private.
Thanks for the report - it's true that error messages in diamond are 
suboptimal when the failure occurs for reasons other than inference.
I'll file a bug on this one.

Maurizio
>
> Steps to reproduce:
>
> I'm using javac version 1.7.0_25.
>
> 1) Create these 2 files.
>
> MyList.java:
> public class MyList<E> {
>   private MyList() { }
> }
>
> Test.java:
> public class Test {
>   public void foo() {
>     MyList<Object> foo = new MyList<>();
>   }
> }
>
> 2) Compile.
>
> $ javac Test.java
> Test.java:3: error: cannot infer type arguments for MyList<>
>     MyList<Object> foo = new MyList<>();
>                          ^
> 1 error
>
> It would be great if javac produced a more helpful error message in 
> this case.
>
> Thanks,
> Eddie