[coverage] Conformance findings: CLOUDFETCH-017

[peco-engineer-bot]

Summary

Surfaced by the multi-language coverage fan-out while conformance-testing these SPEC-IDs against databricks/databricks-sql-python. 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-python) is fixed, then flips green as a tripwire.

Findings

• CLOUDFETCH-017 [sea]: kernel (SEA) retries a persistently failing CloudFetch chunk download without bound — the drain never returns (blocked past 75s after ~7 presigned-URL GETs) instead of charging each completed failure against the download attempt budget and surfacing a terminal error; same non-termination class as databricks-odbc#144
  • failing test: test_persistent_download_failure_is_terminal_and_attempt_bounded (see the coverage PR diff under tests/)

Reproduce & Expected

CLOUDFETCH-017 — A CloudFetch chunk download that fails PERSISTENTLY must become terminal: the driver charges every completed download failure against its download attempt budget, surfaces an error to the consumer, a…

Reproduce:

• Every cloud-storage GET for the chunk fails with HTTP 500 for the whole test.
• Execute a CloudFetch (external-links) query and drain the result. The drain
  must terminate with an error rather than hang; wrap it in a wall-clock bound
  so an unbounded-retry driver fails instead of running to the CI timeout.

Expected (per the shared spec):

• full assertion contract:

result:
- exception_thrown: true
  description: 'The persistent download failure is propagated to the caller as an
    error.

    Message content is not asserted (drivers word cloud-download errors

    differently); the point is that the read terminates in failure, not that

    it returns partial rows or blocks.

    '
- cloud_download_count:
    min: 1
    max: 2
  description: "The failing chunk is downloaded at least once and NO MORE than the\n\
    configured download attempt budget. A driver that skips attempt\naccounting on\
    \ the stale-link path exceeds this bound (or never stops).\nA driver exposing\
    \ no attempt-budget knob asserts its own small fixed\nupper bound instead \u2014\
    \ the contract is that the count is BOUNDED, not\nthat it equals a particular\
    \ number.\n"
- elapsed_seconds_range:
    min: 0
    max: 120
    description: The failing drain terminates well inside the CI timeout.

Context

• The behavior was first fixed in a DIFFERENT driver — reference PR: https://github.com/databricks/databricks-sql-kernel/pull/206 — which seeded the shared language-neutral spec. This issue tracks the same conformance gap in databricks/databricks-sql-python; 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/1136