Use all known Scala versions when -DallVersions set

[findepi]

Relates to #6167

[findepi]

Can these interfere with

iceberg/build.gradle

Lines 1040 to 1066 in 75b3a05

	// The Iceberg-Build builds for only one Scala version at a time, so the BOM would also
	// only contain artifacts for that single Scala version. The following code ensures that
	// the BOM references the artifacts for all Scala versions.
	def sparkScalaPattern = ~"(.*)-([0-9][.][0-9]+)_([0-9][.][0-9]+)"
	def sparkScalaVersions = [
	"3.3": ["2.12", "2.13"],
	"3.4": ["2.12", "2.13"],
	"3.5": ["2.12", "2.13"],
	]
	rootProject.allprojects.forEach {
	// Do not include ':iceberg-spark', the bom itself and the root project.
	if (it.name != 'iceberg-bom' && it != rootProject && it.childProjects.isEmpty()) {
	if (it.name.startsWith("iceberg-spark-")) {
	def sparkScalaMatcher = sparkScalaPattern.matcher(it.name)
	if (!sparkScalaMatcher.find()) {
	throw new GradleException("Expected a Spark/Scala version combination in Gradle project name ${it.name}")
	}
	def prjName = sparkScalaMatcher.group(1)
	def sparkVer = sparkScalaMatcher.group(2)
	for (def scalaVer in sparkScalaVersions[sparkVer]) {
	add("api", "${it.group}:$prjName-${sparkVer}_$scalaVer:${it.version}")
	}
	} else {
	add("api", project(it.path))
	}
	}
	}

?

[nastra]

I think this would interfere with a few other places where allVersions is being used

[findepi]

should we then document why scala is left out and rename allVersions to avoid people assuming something incorrectly?

for exemple one could assume that ./gradlew -DallVersions=true spotlessApply will reformat all of the code, but i don't think it would

[nastra]

./gradlew -DallVersions=true spotlessApply will re-format all the code. Code formatting is independent from the underlying Scala version

[findepi]

thanks for proving my example wrong. i do think the basic idea still holds though. the "all versions" property is not quite "all versions"

[nastra]

the purpose when I added allVersions was to reflect all module versions that we release but I see how this can cause confusion with Scala versions (which only apply for Flink and Spark).

I still don't think we can blindly just include all Scala versions, because they would only apply across Spark modules and will most likely cause issues with Flink modules when you pass Scala 2.13 when building Flink versions.

[findepi]

would it help if we called this allModules?

[nastra]

I'm fine calling this allModules(we just need to make sure we properly reflect this in docs and any other place where allVersions is currently used.

Do you see a specific use case where you'd like to build all modules across all supported scala versions?

[findepi]

For example to check the code compiles correctly, but i understand now this may not be needed and that scala version config is more for CI that for building.
Thanks, will rename to allModules!

[findepi]

#10499