[VL] Re-enable SPARK-36286 timestamp cast test with Velox 18502

[Smallfu666]

What

Re-enables the SPARK-36286: invalid string cast to timestamp test that was disabled across all Spark version shims (spark33/34/35/40/41).

The test was disabled because Velox accepted CAST('2015-03-18T' AS TIMESTAMP) as a valid date with zeroed time (2015-03-18 00:00:00), while Spark returns NULL (a T separator must be followed immediately by the time component).

Velox PR facebookincubator/velox#18502 fixes the parser to reject trailing-T strings. This PR re-enables the Gluten regression test to validate that fix.

Change

• Remove .exclude("SPARK-36286: invalid string cast to timestamp") from VeloxTestSettings.scala in all 5 shims.
• Set UPSTREAM_VELOX_PR_ID="18502" in ep/build-velox/src/get-velox.sh so Gluten CI builds its pinned Velox baseline with the proposed fix applied.

Testing

UPSTREAM_VELOX_PR_ID="18502" is intentionally set so that Gluten CI:

1. checks out its normal pinned Velox baseline;
2. downloads Velox PR #18502 as a patch;
3. applies the patch;
4. builds Gluten/Velox;
5. runs the re-enabled SPARK-36286 test against the actual proposed Velox fix.

Once #18502 is merged and Gluten's pinned Velox baseline contains the fix, the PR ID pin can be removed.

Related

• Velox PR: fix(spark): Reject trailing-T string in Spark cast to timestamp facebookincubator/velox#18502
• Gluten issue: [VL] Fix corner cases for casting varchar to timestamp #8984

[github-actions]

Run Gluten Clickhouse CI on x86

[Copilot]

Pull request overview

This PR re-enables the SPARK-36286 regression test for invalid string-to-timestamp casting in the Velox backend across all Spark shims, and attempts to validate the behavior by applying an upstream Velox PR patch during the Velox build.

Changes:

• Removed the .exclude("SPARK-36286: invalid string cast to timestamp") filter from VeloxTestSettings.scala in Spark 3.3/3.4/3.5/4.0/4.1 shims.
• Set UPSTREAM_VELOX_PR_ID="18502" in ep/build-velox/src/get-velox.sh to auto-apply the upstream Velox fix during CI builds.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
gluten-ut/spark33/src/test/scala/org/apache/gluten/utils/velox/VeloxTestSettings.scala	Re-enables the SPARK-36286 test by removing its exclusion.
gluten-ut/spark34/src/test/scala/org/apache/gluten/utils/velox/VeloxTestSettings.scala	Re-enables the SPARK-36286 test by removing its exclusion.
gluten-ut/spark35/src/test/scala/org/apache/gluten/utils/velox/VeloxTestSettings.scala	Re-enables the SPARK-36286 test by removing its exclusion.
gluten-ut/spark40/src/test/scala/org/apache/gluten/utils/velox/VeloxTestSettings.scala	Re-enables the SPARK-36286 test by removing its exclusion.
gluten-ut/spark41/src/test/scala/org/apache/gluten/utils/velox/VeloxTestSettings.scala	Re-enables the SPARK-36286 test by removing its exclusion.
ep/build-velox/src/get-velox.sh	Pins application of Velox PR #18502 during source preparation.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Smallfu666]

@rui-mo Hi, this is the paired Gluten PR for Velox #18502. CI is green now — could you take a look when you have a chance? Thanks!

[rui-mo]

Thanks @Smallfu666. This change looks good to me. I left a few comments on facebookincubator/velox#18502. Would you please take a look?

Once the Velox PR is merged, Gluten can sync to the updated Velox version, and the PR number override in "get-velox.sh" should no longer be needed.

[Smallfu666]

Thanks @rui-mo — I missed those inline comments earlier. I'm going through them now, especially the whitespace-handling point, and will update #18502 accordingly. Thanks for the heads-up!