rdtr opened a new pull request, #13259: URL: https://github.com/apache/gluten/pull/13259
## What changes are proposed in this pull request? Re-enables 4 of the 6 Spark 4.2 tests in #13178 that #13163 disabled: - `GlutenMathExpressionsSuite` "SPARK-56089: asinh/acosh fdlibm algorithm coverage" and `GlutenMathFunctionsSuite` "acosh": Velox returns NaN as `0xfff8000000000000` (from glibc), while the JVM's `Double.NaN` is `0x7ff8000000000000`. Gluten's result comparison (`GlutenQueryTestUtil.compare` and `GlutenTestsTrait.checkResult`) compares NaN by its raw bits, so the two NaNs did not match. They are both NaN, and Spark's own tests treat any two NaNs as equal since SPARK-55975. The comparison now does the same. This is in `gluten-ut/common`, so it applies to all Spark versions, but it only changes the result when both values are NaN. - `GlutenBroadcastExchangeSuite` "SPARK-56455: broadcast should fail when table size exceeds maxBroadcastTableSize": fixed by #13192. - `GlutenReplaceHashWithSortAggSuite` "Gluten - do not replace hash aggregate if child does not have sort order": since SPARK-55551, a broadcast hash join reports its stream side's partitioning on the build keys too, so the query needs no shuffle before the final aggregate and Gluten merges the partial and final aggregates into one. The test now expects 1 hash aggregate instead of 2. The other 2 tests stay disabled: - `GlutenXmlExpressionsSuite` "TIME type with arrays" fails in Spark's conversion of arrays of `TIME(3)` back to rows, not in Velox. - `GlutenDateExpressionsSuite` "TruncTimestamp of Long.MinValue overflows with ArithmeticException" needs a separate fix and is not covered here. Part of #13178. ## How was this patch tested? The changed tests compile on Spark 4.2 and Spark 3.5. The Spark 4.2 UT CI runs the re-enabled tests. ## 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]
