Refactored buildSrc Gradle-based tests around GradleFixture and reduce resource use.

[AlexeyKuznetsov-DD]

What Does This Do

Shares the TestKit directory used by GradleFixture so the Gradle daemons spawned by buildSrc plugin tests can be reused across test methods. Several of these tests run real Gradle builds, including Kotlin Gradle DSL script compilation, and paying for a fresh TestKit daemon/cache on every test was expensive in check_build_src.

The fixture keeps the shared TestKit directory outside each JUnit @TempDir so temp-dir cleanup does not race daemon file locks, and it stops the TestKit daemons from a JVM shutdown hook. Because shared daemons can also share plugin classloader state, MuzzleMavenRepoUtils no longer caches the MAVEN_REPOSITORY_PROXY-derived repository list in a daemon-static lazy value; it now builds the default muzzle repository list on demand.

The duplicated GradleRunner setup in individual tests has also been consolidated into GradleFixture. Test classes now extend the fixture directly and use focused helpers such as writeFile, writeRootProject, writeSettings, writeGradleProperties, writeJavaSource, addSubproject, dir, file, and buildFile that read more like a DSL than boilerplate.

Motivation

1. Reduce check_build_src overhead. Reusing the TestKit daemon/cache amortizes Gradle and Kotlin DSL setup across the buildSrc plugin tests.

2. Keep daemon reuse correct. Shared daemons make daemon-static state visible across fixture builds, so environment-derived muzzle repository configuration must be recomputed instead of cached.

3. One way to write a fixture test. Centralizing GradleRunner setup and project-file helpers makes the tests shorter and keeps build fixture behavior uniform.

Additional Notes

Stats from GitLab

Job time reduced from ~10mins to ~5mins!

Memory reduced:

**Before:**                                       | **After:**
--------------------------------------------------------------------------------------------------------
RAM memory                          : 194720488   | RAM memory                          : 389949356  !!
cgroup v2 memory.peak               : 19437047808 | cgroup v2 memory.peak               : 7072915456 !!
cgroup v2 memory.max                : 21474836480 | cgroup v2 memory.max                : 21474836480
cgroup v2 memory.high               : max         | cgroup v2 memory.high               : max
cgroup v2 memory.current            : 2583728128  | cgroup v2 memory.current            : 1718136832 !!
cgroup v2 memory.pressure           :             | cgroup v2 memory.pressure           :
 some avg10=0.00 avg60=0.00 avg300=0.00 total=733 |  some avg10=0.00 avg60=0.00 avg300=0.00 total=0  !!
 full avg10=0.00 avg60=0.00 avg300=0.00 total=732 |  full avg10=0.00 avg60=0.00 avg300=0.00 total=0  !!

[bric3]

This change is nice, however I'd rather not switch back to groovy dsl.

Since it looks ai generated for some parts, I'd rather convert the remaining fixture to kotlin dsl.

[bric3]

This switched to groovy dsl.

[AlexeyKuznetsov-DD]

@bric3 thanks for pointing to this!

I switched all scripts to Kotlin DSL. And ./gradlew -PrunBuildSrcTests=true :buildSrc:test started to take ~20min. I investigated and after couple of rounds of refactoring I was able to do more-or less elegant way to reuse Gradle daemons! Now this test suite passed in ~1.5 mins locally and do not consume insane amount of RAM. But the change is quite huge and non trivial, please do one more round of review. With daemons reuse results are very impressive.

[bric3]

Cool thanks for this ! Indeed that's the way to go with Gradle : "daemon reuse". But it may be possible some tests need to offer a way for non reusable daemons for some tests 🤔 , maybe not yet though.

I'll take a second review