Test needs to assume script splitting is enabled

[bitwiseman]

This test failure made it past CI because the project uses the Java version to decide whether to enable script splitting or not. When run outside of CI, that choice is apparently not preserved.

[jglick]

See jenkinsci/bom#451 (comment) and #415 (comment)

[jglick]

Fine as a hotfix if it works. I suspect the basic problem is the abuse of POM profiles to control test execution. (To start with, recall that <activation> is ignored if you explicitly pass a list of profiles on the CLI!) Basically anything “creative” in a POM is a warning sign.

If this script splitting, whatever it is, behaves differently on Java 8 vs. 11, then better to have the test code itself decide whether to run or not. Or if you have some sort of feature flag, then use FlagRule to verify behavior with nondefault settings (which ensures that the flag is reset when the test finishes), and as needed use @Parameterized or other POJO idioms to run analogous tests with the flag both on and off.

[jtnord]

guessing you do not actually want this to be public (except for test code?) so you can change it at will?

Suggested change

	public class RuntimeASTTransformer {
	@Restricted(NoExternalUse.class) // for test use only
	public class RuntimeASTTransformer {

[bitwiseman]

@jtnord
Actually, no. I considered that, but Groovy that is loaded and run in pipeline has difficulty with attributes. Making this public is not great, but not horrible.

[jtnord]

if it works 🤷

[bitwiseman]

Follow on to e1ef424 . We can't sent environment variables in buildPlugin() so that change uses java version to turn on script spitting for at least one run in CI.

This skips this test if script splitting is not enabled - it requires it to work.