HyukjinKwon opened a new pull request, #97:
URL: https://github.com/apache/spark-connect-rust/pull/97
### What changes were proposed in this pull request?
Make the maturin **sdist** contain the repo-root `LICENSE`, and add a
release-CI guard that fails if a declared `License-File` is missing from a
built sdist.
- `pyproject.toml`: add `LICENSE` to `[tool.maturin] include` with `format =
"sdist"` (the wheel already bundles it automatically and is unchanged).
- `.github/workflows/release.yml`: after `maturin sdist` (dry run + RC),
assert every `License-File` declared in `PKG-INFO` is actually present in the
tarball.
### Why are the changes needed?
maturin auto-bundles the license into the **wheel**
(`*.dist-info/licenses/LICENSE`) but does **not** pull the repo-root `LICENSE`
into the **sdist** — the maturin crate lives under `crates/pyspark-rs/`, so a
file at the repository root is outside it and never gets packaged. Both
artifacts still declare `License-File: LICENSE` in their metadata, so PyPI (PEP
639) rejects the sdist:
```
400 License-File LICENSE does not exist in distribution file
pyspark_client_rust-<v>.tar.gz at pyspark_client_rust-<v>/LICENSE
```
while the wheels upload fine. This surfaced only at **finalize of 4.2.0**:
the four wheels reached PyPI but the sdist was rejected, blocking
crates.io/tagging behind it. Neither `twine check` nor `cargo publish
--dry-run` validates PEP 639 license-file presence, so the existing dry runs
stayed green. The 4.2.0 release itself was completed by hand-adding the
`LICENSE` to that sdist; this PR fixes the root cause so future releases
(4.3.0+) build a valid sdist and CI catches the regression.
### Does this PR introduce _any_ user-facing change?
No. The published sdist gains the `LICENSE` file it already promised in
metadata; the wheel is unchanged.
### How was this patch tested?
- Verified the failing sdist and the fix mechanics locally (sdist gains
exactly `pyspark_client_rust-<v>/LICENSE`, all other files byte-identical).
- The new `release.yml` guard exercises the fix on the next scheduled dry
run / RC cut (it builds the sdist and asserts the declared `License-File` is
present).
> Note for the release manager: this project's PR titles carry a
`[SPARK-XXXXX]` JIRA id — please prepend the appropriate JIRA before merging.
This pull request and its description were written by Isaac.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
