GROOVY-8283: field hides getter of super class (not interface)

[eric-milles]

For the dynamic side of 8283, field can be selected in MetaClassImpl. If access method is declared by an interface, the interface method is the one indexed so must be considered pervasive in the type hierarchy (cannot be safely hidden). Not sure if this kind of name hiding should be extended to inner classes; hopefully the commit is enough of a first step to determine if this protocol should be changed or left alone.

@blackdrag @paulk-asert

TODO:

• STC selection of field vs setter (verify what StaticTypeCheckingVisitor does vs MetaClassImpl)
• static compiler classgen -- I think it currently selects the method

https://issues.apache.org/jira/browse/GROOVY-8283

[eric-milles]

@blackdrag @paulk-asert If I refresh the branch, do you think this is viable? The test case should illustrate the new hiding behavior.

[eric-milles]

For circumstances that go through ScriptBytecodeAdapter, the sender class is not propagated to MetaClassImpl. Access checks give different results, since the sender is re-mapped to the object.

• groovy/src/main/java/org/codehaus/groovy/runtime/ScriptBytecodeAdapter.java

  Line 472 in 9819280

  return groovyObject.getProperty(messageName);

• groovy/src/main/java/org/codehaus/groovy/runtime/ScriptBytecodeAdapter.java

  Line 504 in 9819280

  groovyObject.setProperty(messageName, messageArgument);

            package p
            class A {}
            class B {}
            class C {
                boolean setter
                protected A foo = new A()
                A getFooA() { return this.@foo }
                void setFoo(A a) { setter = true; this.@foo = a }
            }
            class D extends C {
                protected B foo = new B() // hides A#foo; should hide A#setFoo in subclasses
                B getFooB() { return this.@foo }
            }

import p.*
            class E extends D {}
            def e = new E()
            e.foo = null // not the field from this perspective
            assert e.setter // fails

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 82.85714% with 6 lines in your changes missing coverage. Please review.

Project coverage is 68.8289%. Comparing base (da495fa) to head (8452e8a).
Report is 6 commits behind head on master.

Files with missing lines	Patch %	Lines
src/main/java/groovy/lang/MetaClassImpl.java	80.0000%	1 Missing and 3 partials ⚠️
...ovy/classgen/asm/sc/StaticTypesCallSiteWriter.java	60.0000%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@                Coverage Diff                 @@
##               master      #1767        +/-   ##
==================================================
+ Coverage     68.8221%   68.8289%   +0.0068%     
- Complexity      29437      29460        +23     
==================================================
  Files            1420       1420                
  Lines          113141     113156        +15     
  Branches        19543      19552         +9     
==================================================
+ Hits            77866      77884        +18     
+ Misses          28735      28731         -4     
- Partials         6540       6541         +1     

Files with missing lines	Coverage Δ
...codehaus/groovy/runtime/ScriptBytecodeAdapter.java	69.9266% <100.0000%> (ø)
...roovy/transform/stc/StaticTypeCheckingVisitor.java	87.7228% <100.0000%> (+0.0422%)	⬆️
...ovy/classgen/asm/sc/StaticTypesCallSiteWriter.java	73.8683% <60.0000%> (-0.1442%)	⬇️
src/main/java/groovy/lang/MetaClassImpl.java	77.7839% <80.0000%> (-0.0369%)	⬇️

... and 5 files with indirect coverage changes