rdtr opened a new pull request, #13263: URL: https://github.com/apache/gluten/pull/13263
## What changes are proposed in this pull request? `velox_backend_ansi.yml` runs the backends-velox and spark-ut tests with ANSI mode on and `spark.gluten.sql.ansiFallback.enabled=false` only for Spark 4.0 and 4.1. This PR adds the same two jobs for Spark 3.5, as discussed in #12892: - `spark-test-backends-velox-ansi-spark35` and `spark-test-spark-ut-ansi-spark35` mirror the Spark 4.0 jobs. They use the container and Python setup of the regular Spark 3.5 jobs in `velox_backend_x86.yml` (`centos-8-jdk8`, Java 17, Scala 2.12). - Spark 3.5 turns ANSI mode off by default, but its default for `spark.sql.ansi.enabled` reads `SPARK_ANSI_SQL_MODE`, which the workflow already sets to `true`. - Like the Spark 4.0 jobs, the Spark 3.5 results are not part of the analysis step, which reads only the Spark 4.1 artifacts. `analyze-only` now skips the Spark 3.5 artifacts as well. The workflow runs only on demand (`/ansi-test`), so this doesn't affect the regular CI. A PR can't run the new jobs before it is merged, so I expect new failures on Spark 3.5. I'll fix them or exclude them with a reason in follow-up PRs, and then enable the ANSI-on cast suite on Spark 3.5. Part of #12892. ## How was this patch tested? Checked that the workflow file is valid YAML and that the new jobs match the Spark 4.0 jobs except for the Spark 3.5 settings. The jobs can only run after this PR is merged. ## Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code (Claude Opus 5.5) -- 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]
