Bug triage results: 2026-06-22

[mbutrovich]

Triage pass over the open requires-triage queue, per the project Bug Triage Guide.

• Total issues processed: 15 (12 triaged, 3 skipped, 0 failed)
• Priority counts applied: priority:critical 1, priority:medium 7, priority:low 4
• Guide: docs/source/contributor-guide/bug_triage.md

Labels have already been applied and requires-triage removed from each issue listed under "Triaged". A reviewer should spot-check the calls and close this issue when satisfied. To correct a label, edit the affected issue directly.

Triaged

priority:critical

• [Bug] make_timestamp does not throw under spark.sql.ansi.enabled=true (#4679)
  • Area labels: area:expressions
  • Rationale: under ANSI mode Spark throws on out-of-range arguments while Comet silently returns NULL, ignoring spark.sql.ansi.enabled; the guide classifies config-ignoring silent divergences as critical (decision-tree step 1).

priority:medium

• [Proposal] Scan I/O acceleration: node-local fragment cache, asynchronous prefetch, and cache-affinity scheduling (#4695)
  • Area labels: area:scan
  • Rationale: net-new performance feature with a workaround today (re-fetch on every scan); missing-capability gap per decision-tree step 3.
• pref: Use builtin compression for arrow ipc writer (#4655)
  • Area labels: area:shuffle
  • Rationale: shuffle write-path performance improvement; current per-batch ZSTD framing works, so it is an optimization with a workaround.
• Add optional native Lance scan support (#4632)
  • Area labels: area:scan
  • Rationale: opt-in new reader; Lance tables already work through Spark's Lance reader, so this is a feature gap with a fallback.
• perf: use aligned slice access in SparkUnsafeArray bulk append (#4626)
  • Area labels: area:shuffle
  • Rationale: native shuffle bulk-append performance work; the existing per-element path is correct, so this is an optimization with a workaround.
• [EPIC] Support Spark interval types (CalendarInterval / YearMonthInterval / DayTimeInterval) and interval expressions (#4540)
  • Area labels: area:expressions
  • Rationale: unsupported types cause fallback to Spark for any interval-carrying query; missing-feature gap with a Spark fallback workaround.
• [EPIC] Provide JVM/codegen-dispatch implementations for Incompatible expressions so they never fall back by default (#4506)
  • Area labels: area:expressions
  • Rationale: feature work to keep Incompatible expressions native via codegen dispatch instead of falling back; functional gap with a workaround.
• Gate non-default StringTypeWithCollation inputs on Spark 4.0 datetime expressions (#4646)
  • Area labels: area:expressions
  • Rationale: non-default collations are silently treated as compatible and run natively; reporter tracked this as medium (split from date/time expression audit follow-ups (from #4448) #4502). See escalation note below.

priority:low

• Add CONTRIBUTING.md with link to the contributor guide (#4694)
  • Area labels: none
  • Rationale: documentation/tooling task with no runtime impact (decision-tree step 4).
• Move static support decisions from serde convert into getSupportLevel (#4673)
  • Area labels: area:expressions
  • Rationale: behavior-preserving internal refactor of the serde framework; no user-facing effect (decision-tree step 4).
• Discussion: remove incompatible native implementations that have a codegen dispatch fallback? (#4654)
  • Area labels: area:expressions
  • Rationale: design discussion, no functional bug or regression; lowest bucket per decision-tree step 4.
• Discussion: Should Comet add geospatial (ST_*) function support? (#4455)
  • Area labels: none
  • Rationale: scope/design discussion, no functional bug; lowest bucket per decision-tree step 4.

Escalations to consider

• Gate non-default StringTypeWithCollation inputs on Spark 4.0 datetime expressions (#4646)
  • Today these datetime expressions run natively for non-default collations where results "may diverge" from Spark, which matches the guide's decision-tree step 1 ("Can this bug cause silent wrong results?") and could justify priority:critical. Left at priority:medium because the reporter and date/time expression audit follow-ups (from #4448) #4502 tracked it as medium and the divergence only occurs with non-default collations on Spark 4.0; flagging for reviewer judgment.

Skipped - needs more info

• Bug triage results: 2026-06-11 (#4625)
  • Prior triage summary issue, not a bug; left requires-triage in place for the reviewer to close.
• Bug triage results: 2026-06-01 (#4548)
  • Prior triage summary issue, not a bug; left requires-triage in place for the reviewer to close.
• Bug triage results: 2026-05-26 (#4441)
  • Prior triage summary issue, not a bug; left requires-triage in place for the reviewer to close.

Notes

• The guide lists spark 4 as a pre-existing area indicator, but no spark 4 label currently exists in the repo (only spark sql tests does). Gate non-default StringTypeWithCollation inputs on Spark 4.0 datetime expressions #4646 is Spark-4.0-specific and would otherwise have received it; left off and noted here rather than creating a new label.