refactor scalac java_binary target creation

[gergelyfabian]

I'd like to remove Scala compiler extensions from the Scala toolchain, that I could be potentially building myself (in the same monorepo, build the compiler extension, but also use it).

Till now the only way to add Scala compiler dependencies was a toolchain.
If I left the compiler extension as a toolchain dependency there, I could not build the compiler extension due to a cycle in deps (building the compiler extension would depend on itself through the toolchain).

I'd like to work this around by defining my own scalac java_binary target, where I can define my own dependencies (and then create rule names with macros that override the scalac target).

To do this without duplicating code from rules_scala, refactor SCALAC_DEPS parameter and scalac target definitions to a Starlark file to make them reusable for outside users.

Added:

• define_scalac()
• define_scalac_bootstrap()

A user could do:

EXTENDED_SCALAC_DEPS = DEFAULT_SCALAC_DEPS + [
    "@//my/compiler/dependency:library",
]

define_scalac(
    name = "scalac_with_extensions",
    deps = EXTENDED_SCALAC_DEPS,
)

And then build that compiler library in the same monorepo, this way, providing compiler dependencies without toolchains.

See #1674 for more information.

[gergelyfabian]

Rebased with latest master to ensure the build for latest Bazel does not fail.

[gergelyfabian]

In fact, I could move the java_binary's definition to the Starlark file too...

[gergelyfabian]

In fact, I could move the java_binary's definition to the Starlark file too...

Just did that.

[gergelyfabian]

Moved the scalac target definitions to Starlark too, so that those are also reusable. Implemented the idea with Label from @mbland (Thanks!) and added a more detailed description. This could be a user-friendly solution to #1674 in fact.

[gergelyfabian]

Not sure why the build failed, I just rebased with latest master.

[mbland]

This looks like the culprit: https://buildkite.com/bazel/rules-scala-scala/builds/5293#0194694d-dea6-481f-85ca-36bcc3da80d3/126-462

ERROR: /workdir/test/gen_src/BUILD:20:13: scala @//test/gen_src:uses_gen_file failed:
  (Exit 1): scalac failed: error executing command
  (from target //test/gen_src:uses_gen_file)
  bazel-out/k8-opt-exec-2B5CBBC6/bin/src/java/io/bazel/rulesscala/scalac/scalac
  @bazel-out/k8-fastbuild/bin/test/gen_src/uses_gen_file.jar-0.params

bazel-out/k8-fastbuild/bin/test/gen_src/foo.scala:1: error: Invalid literal number

BTW, the other ERROR further down about a nonexistent //test/coverage/... package I've fixed in one of the commits in #1678. That "fix" doesn't actually affect the test result, because both of the builds launched by the script fail in the same way, but it no longer hides the underlying breakage.

The error from this PR seems to be a different issue altogether. I'm not immediately sure what it is.

[gergelyfabian]

The error from this PR seems to be a different issue altogether. I'm not immediately sure what it is.

It seems it's non-deterministic (and probably not related to this PR). Amended my commit, force-pushed, and the rerun of the same test passed.

[simuons]

Thanks, @gergelyfabian! Sorry for unreasonable delay,