rdtr opened a new pull request, #13592: URL: https://github.com/apache/gluten/pull/13592
## What changes were proposed in this pull request? Related to #10134, requested in #13263: add a test tag for suites whose results depend on ANSI mode, and make the ANSI workflow run only those. - New tag `org.apache.gluten.tags.AnsiTest` in `gluten-substrait/src/test/java` (the test-jar both `backends-velox` and `gluten-ut` already depend on). - `velox_backend_ansi.yml`: the Spark 4.0 / 4.1 jobs pass `-DtagsToInclude=org.apache.gluten.tags.AnsiTest`. The backends-velox jobs also pass `-Dtest=none -DfailIfNoTests=false -Dsurefire.failIfNoSpecifiedTests=false` so surefire's Java tests are skipped like in the spark-ut jobs. - `org.apache.spark.tags.ExtendedSQLTest` is still excluded, so `GlutenSQLQueryTestSuite` stays out of the ANSI jobs, as before. - Regular CI is unchanged: no other workflow or pom passes `tagsToInclude`, and the annotation has no effect without it. ### Tagged suites backends-velox: | Suite | Why | |---|---| | `ArithmeticAnsiValidateSuite` | overflow / divide-by-zero with ANSI on and off | | `MathFunctionsValidateSuite`, `MathFunctionsValidateSuiteAnsiOn` | math functions with ANSI-dependent errors | | `ScalarFunctionsValidateSuite` | ANSI-dependent scalar functions | | `DateFunctionsValidateSuite` | date/time functions | | `VeloxAggregateFunctionsDefaultSuite`, `VeloxAggregateFunctionsFlushSuite` | aggregate overflow (sum/avg) | gluten-ut (Spark 4.0 and 4.1): | Suite | Why | |---|---| | `GlutenCastWithAnsiOnSuite`, `GlutenCastWithAnsiOffSuite`, `GlutenTryCastSuite` | cast | | `GlutenArithmeticExpressionSuite`, `GlutenIntervalExpressionsSuite`, `GlutenMathExpressionsSuite`, `GlutenMathFunctionsSuite` | arithmetic overflow | | `GlutenDateExpressionsSuite`, `GlutenDateFunctionsSuite` | date/time | | `GlutenCollectionExpressionsSuite` | collection functions (index out of bounds) | | `GlutenDataFrameAggregateSuite` | aggregates | | `GlutenTryEvalSuite` | `try_*` | | `GlutenQueryExecutionAnsiErrorsSuite`, `GlutenQueryExecutionErrorsSuite` | error conditions | ### Possible follow-up `GlutenSQLQueryTestSuite` (SQL golden files with ANSI variants) is not in this set because it is an `ExtendedSQLTest` and the ANSI jobs have never run it. In a local Spark 4.0 run it adds 705 tests: 170 passed, 44 failed, 491 ignored. It could be added separately. ## How was this patch tested? Spark 4.0, `-Pspark-ut`, `SPARK_ANSI_SQL_MODE=true SPARK_TESTING=true`, with the workflow's flags (`-DtagsToInclude=org.apache.gluten.tags.AnsiTest`, `ExtendedSQLTest` excluded), `gluten-ut/spark40` only: - 14 suites ran tests, 824 tests: 716 succeeded, 63 failed, 45 ignored. The 63 failures are ANSI-mode differences that remain after tagging (`GlutenArithmeticExpressionSuite` 19, `GlutenCastWithAnsiOnSuite` 24, `GlutenDateExpressionsSuite` 6, and a few in each of the other suites); these jobs exist to report them. The jobs' "Parse test results" steps are unchanged, so the result parsing works as before. - Before: the ANSI workflow is comment-triggered and has no run on main. The last runs that executed the test jobs are from September, and their logs are expired, so I have no before numbers for the four jobs. Without the tag filter the spark40 module has 535 suites. - After: Spark 4.0 as above. I'll add the Spark 4.1 numbers from this PR's `/ansi-test` run. - backends-velox: test sources compile; the tagged run was not executed locally. - `spotless:check` passes. ## 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]
