Wrong completion for constructor code on a generic class

[mauromol]

This was the old GRECLIPSE-1761:

Consider the following Java class:

package m; public class JavaClass<T extends Number> { }

And the following Groovy class:

package m 

class M { 
	JavaClass<Integer> foo = new JavaCl| 
} 

Invoke code assist at "|" and choose the first suggestion.
The expected behaviour is to get either new JavaClass<Integer>() or new JavaClass<>().
What I get instead is: new JavaClass<Number>().

[eric-milles]

Is there a completion setting that must be enabled to get the generics in the completion proposal? All I get is JavaClass() for the completion. This matches what I would expect since the generics need not be given in Groovy.

[mauromol]

I was not aware of this, but after you told me, I discovered that this behaviour happens when the following option is enabled:
Preferences | Java | Editor | Content Assist | Fill method arguments and show guessed arguments
(with either of the radio buttons selected)

I have this option enabled in my workspaces, I can't remember whether I enabled it explicitly or rather it is the default value.

By the way, getting new JavaClass() is OK for me, along with new JavaClass<>() and new JavaClass<Integer>().

[eric-milles]

Thanks. That helps in recreation. Do you actually want the generics for the constructor? Or would a completion of new JavaClass() be acceptable?

[mauromol]

I would say new JavaClass() is perfectly acceptable, I can't think of any side effect (on subsequent type inference or such), as it should be roughly equivalent to new JavaClass<>().

[mauromol]

In 2.9.2.xx-201710051703-e46 this works correctly, thank you! 👍