[coverage] Conformance findings: CONCURRENT-006

[peco-engineer-bot]

Summary

Surfaced by the multi-language coverage fan-out while conformance-testing these SPEC-IDs against databricks/databricks-sql-go. Each finding is committed as an expected-failure (xfail) test in the coverage PR — the test asserts the CORRECT (post-fix) behavior and stays red until THIS driver (databricks/databricks-sql-go) is fixed, then flips green as a tripwire.

Findings

• CONCURRENT-006 [thrift]: Cancelling the query context does not unblock a result-batch fetch already parked inside a CloudFetch cloud-storage download: the fetch stays blocked for the full download (>10s past the interrupt, against a 30s injected delay) instead of racing the download against the cancelled context, so a consumer draining results on a background goroutine has no bounded teardown.
  • failing test: TestInterruptInFlightResultBatchFetch (see the coverage PR diff under tests/)

Reproduce & Expected

CONCURRENT-006 — Verify that a result-stream fetch that is BLOCKED inside a slow batch fetch can be interrupted from another thread, so the blocked fetch call returns PROMPTLY with a cancellation outcome instead of w…

Reproduce:

• Each cloud-storage download stalls 30s, so the drain parks inside it.
• Execute a query whose result is delivered via CloudFetch external links, and obtain
  the interrupt handle for its result stream while the stream is live.
• Drain the result on a background thread/task. It blocks inside the delayed download.
• Long enough that the background fetch is inside the delayed download.
• Fire the result-stream interrupt from the main thread while the fetch is in flight,
  then fire it a SECOND time to exercise idempotency.
• The parked fetch call must return; measure elapsed from the interrupt.
• Fetch once more on the SAME result — must return promptly, no rows.
• Close/dispose the result and release the interrupt handle (either order) after the
  cancel.

Expected (per the shared spec):

• completes without an exception
• result has exactly 0 row(s)
• result has exactly 0 row(s)
• completes without an exception
• full assertion contract:

result:
- label: interrupt_fired
  no_exception: true
- label: interrupt_fired
  elapsed_seconds_range:
    max: 2
- label: fetch_interrupted
  elapsed_seconds_range:
    max: 10
- label: fetch_interrupted
  reports_cancellation: true
- label: fetch_interrupted
  row_count: 0
- label: repeat_fetch
  elapsed_seconds_range:
    max: 5
- label: repeat_fetch
  row_count: 0
- label: teardown
  no_exception: true

Context

• The behavior was first fixed in a DIFFERENT driver — reference PR: https://github.com/databricks/databricks-sql-kernel/pull/312 — which seeded the shared language-neutral spec. This issue tracks the same conformance gap in databricks/databricks-sql-go; the reference PR is for cross-referencing the intended behavior, NOT a change to this repo.
• Coverage PR carrying the reproducing xfail test(s): https://github.com/databricks/databricks-driver-test/pull/1534