Regression from 1.3: compilation fails with “current scope already contains a variable of the name”

[cspurk]

Describe the bug

Compiling a simple Spock specification which works with Spock 1.3 fails with 2.0 and 2.1-SNAPSHOT.

To reproduce

The following specification can be successfully compiled with Spock 1.3 but fails to compile with newer Spock versions (tested with both Groovy 2.5.14 and Groovy 3.0.8):

import spock.lang.Specification

class MySpec extends Specification {

    def foo1() {
        expect:
        'whatever'.with { foo -> !foo.isEmpty() }
        def foo = null
        !foo

        cleanup:
        'irrelevant'.toString()
    }

    def foo2() {
        when:
        'whatever'.with { foo -> !foo.isEmpty() }
        def foo = null

        then:
        notThrown(Exception)
    }
}

The issue doesn’t occur if there is no notThrown(…), no thrown(…) or no cleanup section.

Expected behavior

The specification can be compiled sucessfully.

Actual behavior

The compiler throws the following errors (for Spock 2.+):

startup failed:
/tmp/foo/src/test/groovy/MySpec.groovy: 7: The current scope already contains a variable of the name foo
 @ line 7, column 25.
           'whatever'.with { foo -> !foo.isEmpty() }
                           ^

/tmp/foo/src/test/groovy/MySpec.groovy: 17: The current scope already contains a variable of the name foo
 @ line 17, column 25.
           'whatever'.with { foo -> !foo.isEmpty() }
                           ^

Environment:

Java/JDK

• openjdk 11.0.11
• Groovy 2.5.14 or Groovy 3.0.8

Build tool version

Gradle 7.1.1

Operating System

Ubuntu 20.04.2 LTS

Build-tool dependencies used

Complete build.gradle (tried with different versions as indicated above):

plugins {
    id 'groovy'
}

repositories {
    mavenCentral()
    maven {
        url = 'https://oss.sonatype.org/content/repositories/snapshots'
    }
}

dependencies {
    testImplementation 'org.codehaus.groovy:groovy:2.5.14'
    testImplementation 'org.spockframework:spock-core:2.1-groovy-2.5-SNAPSHOT'
}

[leonard84]

This is the same root cause as #1266 see my comment but this will be trickier to fix. From the user perspective it looks like a regression, but it actually just got more strict with its validations.

I'll have to look at it, for now I would suggest to simply use different names.