ulysses-you opened a new pull request, #58586: URL: https://github.com/apache/spark/pull/58586
### What changes were proposed in this pull request? Update the exchange-count pin in the SPARK-59050 test "genuine unknown keys survive an inner join on a key subset" from 4 to 3, and spell out which exchanges remain. ### Why are the changes needed? SPARK-59248 (#58522) broke this test on master. In that query, `x`'s partition key `k` is column-pruned from its scan output (the join matches on `id` alone). Before SPARK-59248, the pruned key dropped `x`'s whole key-grouped partitioning report, so `x`'s side of the `r JOIN x` join shuffled on its own; with SPARK-59248 the surviving `(id)` projection pairs with `r`'s one-side keyed shuffle and `x`'s exchange disappears. The test's answer check still passes, including the genuine unknown-key row `(1, 'zz')`; only the count was stale. The suite is `@ExtendedSQLTest`, and the "sql - extended tests" job that failed on master does not run on PR CI, which is why #58522 itself was green. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? The updated test, plus the full `KeyGroupedPartitioningSuite` locally (172/172). ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code (GLM 5.3) -- 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]
