andygrove opened a new pull request, #369: URL: https://github.com/apache/datafusion-spark/pull/369
## Which issue does this PR close? Follow-up to #54 and #368. ## Rationale for this change The PySpark runner only ran by hand, so a new `.slt` record whose expected result differs from Spark could merge unnoticed. Now that #368 marks every known difference with `# xfail spark:`, the runner exits 0 on `main` and can gate PRs. ## What changes are included in this PR? - A `spark` job in `pyspark-runner.yml` that installs Java 17 and PySpark 4.2.0 and runs `pyspark_slt_runner.py` over all `.slt` files. The runner already exits 1 when a record differs from Spark without an `# xfail spark:` comment, when a marked record now passes (so stale markers are removed), or when the runner itself errors, and exits 2 if Spark fails. The job fails in any of these cases. - The workflow now also triggers on changes under `crates/sqllogictest/test_files/`. - The scripts README says the runner runs in CI instead of being manual. ## Are these changes tested? Ran the runner locally on `main` with PySpark 4.2.0: 3755 records, 2907 pass, 260 expected failures, 588 skipped, exit 0, about 45 seconds. The CI run on this PR exercises the new job. ## Are there any user-facing changes? No. Contributors who add or change an `.slt` record that differs from Spark need to fix it or mark it with `# xfail spark: <reason>`. -- 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]
