felipepessoto opened a new pull request, #12808: URL: https://github.com/apache/gluten/pull/12808
> **Do not merge.** This pins the native build to an unmerged upstream pull request. It exists to produce a CI result, not to land. ## What this is This is #12783 plus **one line**: it sets `UPSTREAM_VELOX_PR_ID="18536"` in `ep/build-velox/src/get-velox.sh`, so `get-velox.sh` applies the upstream Velox fix before the native build. | | change | Delta suite expected | |---|---|---| | #12783 | enable Velox per-operator output validation | 🔴 **red** | | this PR | same, **+ the upstream Velox fix** | 🟢 **green** | Both are branched from the same commit on `main`, so the fix is the only difference between them. ## Why #12377 has been hard to act on because the failure is intermittent and reports a meaningless row index. Root cause: on a Delta deletion-vector write the scan projects only synthesized columns with a pushed-down filter, and Velox emits a `RowVector` whose row-index child has no rows. That child is wrapped in a dictionary and read out of bounds, so the "row index" is whatever heap memory follows. * Velox issue: https://github.com/facebookincubator/velox/issues/18535 * Velox fix: https://github.com/facebookincubator/velox/pull/18536 Red on #12783 and green here is the before/after evidence, available now rather than after the upstream fix merges and a Velox bump lands. ## Notes * `--build_tests=OFF` in the CI native build, so the patch's `TableScanTest.cpp` hunk is applied but never compiled; only the fix is. * The patch was verified to apply cleanly to the pinned Velox revision, which is what `get-velox.sh` checks with `git apply --check` before applying. * Once the fix is merged and picked up by a Velox bump, both this branch and the config in #12783 should be dropped. Was this patch authored or co-authored using generative AI tooling? Generated-by: GitHub Copilot CLI (Claude Opus 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]
