feat(interpreter): add current_py_toolchain rule for Make variable support

[ctcjab]

Summary

Closes #896.

Adds a current_py_toolchain rule that resolves the active Python toolchain and exposes $(PYTHON3) and $(PYTHON3_ROOTPATH) Make variables via TemplateVariableInfo. This enables bazel_env and genrule integration without depending on rules_python.

• Auto-instantiated in the python_interpreters hub repo at @python_interpreters//:current_py_toolchain
• Also exported from @aspect_rules_py//py:defs.bzl and @aspect_rules_py//py:current_py_toolchain.bzl for standalone use
• Handles both hermetic (PBS) and local system interpreters
• Includes e2e test verifying $(PYTHON3) expansion via genrule

Usage

# With bazel_env:
bazel_env(
    name = "bazel_env",
    toolchains = {
        "python": "@python_interpreters//:current_py_toolchain",
    },
    tools = {
        "python": "$(PYTHON3)",
    },
)

# With genrule:
genrule(
    name = "run_python",
    outs = ["output.txt"],
    cmd = "$(PYTHON3) -c 'print(42)' > $@",
    toolchains = ["@python_interpreters//:current_py_toolchain"],
)

Test plan

• e2e test //e2e/cases/current-py-toolchain-896:test passes
• Existing e2e tests unaffected
• buildifier and buildifier-lint pass

🤖 Generated with Claude Code

[aspect-workflows]

Bazel 8 (Test)

All tests were cache hits

123 tests (100.0%) were fully cached saving 60s.

---

Bazel 9 (Test)

All tests were cache hits

122 tests (100.0%) were fully cached saving 1m 5s.

---

Bazel 8 (Test)

e2e

All tests were cache hits

59 tests (100.0%) were fully cached saving 54s.

---

Bazel 9 (Test)

e2e

All tests were cache hits

59 tests (100.0%) were fully cached saving 1m 3s.

---

Bazel 8 (Test)

examples/uv_pip_compile

All tests were cache hits

1 test (100.0%) was fully cached saving 444ms.

---

Buildifier

[CLAassistant]

All committers have signed the CLA.

[ctcjab]

The pre-commit / conclusion GitHub Actions failures are unrelated to this PR — the workflow's checkout step tries to fetch refs/heads/current-py-toolchain-896 from aspect-build/rules_py, but the branch lives on the chicagotrading fork. It fails with git fetch ... exit code 1 after 3 retries.

The Buildkite CI (which does handle fork PRs correctly) is the relevant check here — it passed on the previous push after the bzl_library fix.

[arrdem]

https://github.com/bazel-contrib/rules_python/blob/d0b9baadc43eff92b3b9df0a04343e7ca653c44a/python/current_py_toolchain.bzl#L47 Hm. This is how rules_python does it too, which I don't love. I thought there was a way to associate the make vars directly with the toolchain type rather than having to have a wrapper rule which generates the variable info.

[ctcjab]

There was a fix in Bazel 8.3.0 (bazelbuild/bazel@0e876b1, referenced in bazelbuild/bazel#14009) that allows genrule.toolchains to accept toolchain_type targets directly. However, looking at the implementation, it's genrule-specific — it explicitly gates on !ruleClass.isStarlark() && ruleClass.getName().equals("genrule"), so custom Starlark rules like bazel_env can't benefit from it.

The wrapper rule pattern (as used by rules_java's current_java_runtime and rules_python's current_py_toolchain) still seems to be the only option for the general case.

[ctcjab]

In light of the above, is it worth merging this, and if Bazel ever provides something better rules_py can take advantage of that then? @arrdem @gregmagolan et al.

[gregmagolan]

cc @xangcastle

[ctcjab]

Friendly bump (since I opened this a month ago and am hoping to avoid acquiring a merge conflict) - thanks in advance for any attention you can give this🙏

[jbedard]

@codex

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 691ec6b122

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ctcjab]

Thanks for the review @jbedard! I've addressed your feedback:

• Using the existing PY_TOOLCHAIN constant from types.bzl
• Added fail() for the case where py3_runtime has neither interpreter nor interpreter_path
• Fixed the e2e test to use the correct ${TEST_SRCDIR}/_main/cases/... runfiles path pattern

Also rebased on latest main. CI is passing — would appreciate another look when you get a chance.