Make smoke tests capability probes; fix 5 that crash on engine divergence

[eerxuan]

Summary

Smoke tests are meant to answer "does the engine support this feature?" — one minimal probe per feature. Several were written as full round-trip assertions, so they failed on cosmetic divergences (e.g. ok returned as int32 vs the double the expected value used) even though the feature works, or crashed with unhelpful TypeError/IndexError when a setup step behaved differently.

Framework (framework/assertions.py)

• assertCommandSupported(result) — passes if the command returned a result rather than erroring; no value/BSON-type comparison. The capability probe.
• assertChangeStreamEvent(result) — command succeeded and cursor.nextBatch is non-empty (an event was emitted); no comparison of event contents.

Fixes (5 that errored on non-matching behaviour)

• getMore / killCursors: extract the cursor id defensively so a failed setup find reports cleanly through the single assertion instead of raising 'OperationFailure' object is not subscriptable.
• changeStream createIndexes / dropIndexes / modify: extract the cursor id defensively and assert an event was emitted, instead of indexing nextBatch[0] into an IndexError.

Reduced to capability probes (25)

index create/properties/types, minN/maxN accumulator+expression+window, $rank, planCacheListFilters, and the collStats/connectionStatus/dataSize/dbStats/hostInfo/logout diagnostics. These now assert only that the command is supported; exact response shape stays the job of each feature's detailed tests.

Each test keeps exactly one assertion (test-format validator compliant).

Not changed here

The assert-on-error smoke tests (commitTransaction, abortTransaction, compactStructuredEncryptionData, setIndexCommitQuorum, authenticate). Asserting a specific error is a poor support signal and can't be special-cased per engine; converting them to happy-case probes is a follow-up.

Testing

Verified against a live engine: 24 non-replica-set fixed smoke tests pass; the 3 $setWindowFields ones pass with the window-fields feature flag on; the 3 changeStream ones are replica_set-gated.

[documentdb-triage-tool]

🤖 Auto-triaged by documentdb-triage-tool.

Applied: compatibility test, enhancement
Project fields suggested: Component test-coverage · Priority P0 · Effort XL · Status Needs Review
Confidence: 0.82 (mixed)

Reasoning

component from path globs (test-coverage, test-framework); P0 keyword "crash" in title; effort from diff stats (129+175 LOC, 31 files); LLM: Adds new assertion helpers and refactors ~30 smoke tests to be defensive capability probes, touching framework/assertions.py and multiple test files across one component.

If a label is wrong, remove it manually and ping @patty-chow so the rules can be tuned. The bot will not re-label items that already have component labels.