Summary
During review of issue #100, we noted that exact-zero hint coverage remains light for the new bounded count/probe helpers:
CountSearchReferencesCountCallersCountCallees
Common exact-zero tests and existing E2E coverage passed, so this is not a correctness bug today. But the helper-specific behaviors introduced for exact-zero hint cost control should have dedicated regression coverage.
Why this matters
The exact-zero flow now depends on separate existence, bounded count, and capped sample probes. Regressions here could silently reintroduce:
- count paths that ignore the caller's
--limit
- sample probes that exceed the 5-row cap
- unnecessary sample queries when the relaxed path is still empty
Proposed work
Add focused CLI and/or MCP regression tests that verify, for references, callers, and callees:
- relaxed count respects the requested limit
- sample lists are capped at 5 entries
- sample probes are skipped when the relaxed query still returns zero hits
Context
Observed during review of the issue #100 exact-zero hint changes.
Summary
During review of issue #100, we noted that exact-zero hint coverage remains light for the new bounded count/probe helpers:
CountSearchReferencesCountCallersCountCalleesCommon exact-zero tests and existing E2E coverage passed, so this is not a correctness bug today. But the helper-specific behaviors introduced for exact-zero hint cost control should have dedicated regression coverage.
Why this matters
The exact-zero flow now depends on separate existence, bounded count, and capped sample probes. Regressions here could silently reintroduce:
--limitProposed work
Add focused CLI and/or MCP regression tests that verify, for
references,callers, andcallees:Context
Observed during review of the issue #100 exact-zero hint changes.