Groovy with method bug fix in Spock 2.4-M7

[lukasgraef]

Describe the bug

I've stumbled over a possible regression in Spock 2.4-M7 when using Groovy's with{}-closure within a then: or expect: block.

The following two Reproducer tests both fail on M7 while both tests are working on M6.

To Reproduce

import spock.lang.Specification

class ReproducerSpec extends Specification {
    def "assert true in closure"() {
        expect:
        "foo".with {
            assert true
        }
    }

    def "return false in closure"() {
        expect:
        "bar".with {
            false
        }
    }
}

Expected behavior

I'm not sure whether the example above is a regression, because I think the behaviour in M7 (asserting the return value of with{}) seems more consistent than the previous behaviour. I just wanted to report this, as it breaks existing test cases, so this new behaviour should at least be clearly documented.

Actual behavior

Java version

Java 17

Buildtool version

Gradle 8.13

What operating system are you using

Windows

Dependencies

org.spockframework:spock-core:2.4-M7-groovy-3.0

Additional context

No response

[lukasgraef]

@AndreasTu
FYI

[Vampire]

This actually just was a bug-fix.
Spock behaved wrong, as it confused the GDK with method with our own with method and did then some rewriting, but also just half rewriting.
If you for example have

expect:
assert null.with {
    false
}

the test also was successful as the rewriting did remove the assert but did not add our power assertion code, so no assertion was checked here.

The two cases you listed are like you said, previously handled wrong by Spock,
as Spock should check the top-level expression as implicit assertion and did not, as for the first the value of the expression is null, for the second false.

If anything breaks due to that change, it was because the tests relied on this bug which was not a good idea in the first place I'd say.
This PR by me fixed those cases and some others: #2162

Not sure whether we should list the fixing of a bug in Spock as breaking change if it only breaks code that relied on a bug. 🤷‍♂️

[AndreasTu]

Not sure whether we should list the fixing of a bug in Spock as breaking change if it only breaks code that relied on a bug.

I would say we should not list it as breaking change.
But maybe add a note to the release notes, that this bug was now fixed regarding the with method, and tests relying on the spock bug may break.
@Vampire what do you think?

[Vampire]

I don't have too strong feelings about it.
But sounds good to me. :-)

[AndreasTu]

I have created the PR #2270 to update the ReleaseNotes.