Reference to field or property with explicit this in then block

[eric-milles]

Describe the bug

Making a reference to the spec's field or property using explicit-this in the then block produces an STC error. No error shows for this.myField in the when block.

this.myField in the then is written as $spock_valueRecorder.record($spock_valueRecorder.startRecordingValue(0), this).myField in the feature method. record returns Object, so as far as STC is concerned, it can't find the property "myField" in java.lang.Object.

Spock 2.4
Groovy 5.0.5

To Reproduce

@groovy.transform.TypeChecked
class SomeSpec extends Specification {
  String myField
  void testField() {
   when:
    this.myField = 'test'
   then:
    this.myField == 'test' // [Static type checking] - No such property: myField for class: java.lang.Object
  }
}

Expected behavior

this.myField checks properly or is exempted from STC in when, then, expect, etc.

Actual behavior

"[Static type checking] - No such property: myField for class: java.lang.Object"

Java version

Java 25.0.2

Buildtool version

Eclipse 2026-06 / Groovy 5.0.5

What operating system are you using

Windows

[Vampire]

I don't think we support @TypeChecked or @CompileStatic but afair too much rely on Groovy duck-typing for this to work.
Otherwise, I guess we probably need to sprinkle in type-casts in various places or make some methods generic instead of Object.

[eric-milles]

Not sure how record works, but if it was T <T> record(int i, T t) instead of Object record(int i, Object o) then you should get inferencing. Also, I'd need to look further to understand why the when block is okay. Might be a source position thing.

[Vampire]

That's what I meant with "make some methods generic instead of Object", yes.
But I somehow doubt that the record method is the only problematic part.

In the when it is not a problem because it is not rewritten to be an assertion.
It is just an assignment and stays like that.
If you in the when write assert this.myField == 'test' you get the same problem there.

I'm not saying we should not make the record generic.
I'm just saying that there will most probably quite some other parts that are not compatible with STC.