[jira] [Created] (GROOVY-8028) @CompileStatic causes behavioral change in how interfaces work

["Chris Erickson (JIRA)" <ji...@apache.org>]

Chris Erickson created GROOVY-8028:
--------------------------------------

             Summary: @CompileStatic causes behavioral change in how interfaces work
                 Key: GROOVY-8028
                 URL: https://issues.apache.org/jira/browse/GROOVY-8028
             Project: Groovy
          Issue Type: Bug
    Affects Versions: 2.4.4
         Environment: Windows / Grails 2.5.4
            Reporter: Chris Erickson


Given an interface:
{code}
interface SomeInterface {
    String name
}
{code}

and a class:
{code}
class SomeClass implements SomeInterface {
    String name
}
{code}

and a wrapper class, for which we have a statically compiled and non-statically compiled version:
{code}
class WrapperClass {
    SomeInterface delegate

    String getName() { delegate.name }
}

@CompileStatic
class StaticCompileWrapperClass {
    SomeInterface delegate

    String getName() { delegate.name }
}
{code}

The behavior of the two classes is different, with no compiler warnings or errors, as demonstrated by this script:

{code}
    def wrapped = new WrapperClass(delegate: new SomeClass(name: "Name"))
    assert wrapped.name == "Name" // ok
    assert wrapped.getName() == "Name" // ok

    def wrapped2 = new StaticCompileWrapperClass(delegate: new SomeClass(name: "Name"))
    assert wrapped2.name == "Name" // fail
    assert wrapped2.getName() == "Name" // fail
{code}

It was pointed out when I reported this to grails (I thought it was a @GrailsCompileStatic issue but it turns out I had defined it differently in my interface) that the interface definition is incorrect (I'd assumed you could define a property on an interface, which would imply the getters/setters) and that leads to the issue.

However, I think that the functional behavior of a class should not change based on a @CompileStatic annotation. Minimally, if it does change, there should be compiler warnings or errors.  The fix may be that the compiler should fail saying that "SomeClass" doesn't implement the interface, or that when it isn't compiled static it also returns null. I could be convinced either way, the inconsistency is my complaint.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)