Repository navigation
[EPIC] Fix Spark SQL failures when shuffle mode is jvm #1256
Description
Describe the bug
When we run Spark SQL tests with shuffle mode of jvm we see some test failures.
- count distinct on NaN produces incorrect results #1238
- Invalid argument error: Invalid arithmetic operation: Int32 - Int64 #1246
- Spark SQL test "window function: partition and order expressions" fails with scala.MatchError #1249
- Spark SQL test "write null values" fails with Arrow error "Accounted size went negative" #1250
- Comet Internal Error: Output column count mismatch: expected 0, got 1 #1251
- Spark SQL test "test with low buffer spill threshold" fails when shuffle mode is jvm #1252
Steps to reproduce
No response
Expected behavior
No response
Additional context
No response
Activity