Parametrise the skip-reason emptiness check over the real derivation

[WilfordGrimley]

The finding

test_no_declared_value_is_empty_or_whitespace parametrised over EXPECTED_SKIP_REASONS — the hand-written dict 100 lines above it — and asserted those literals were non-empty and stripped. That is an assertion about text typed into the test file. No implementation change can falsify it.

Mutation, verified: set EXTRACTOR_FETCH_FAILED_SKIP_REASON = " " in image_evidence.py. A whitespace-only reason is invisible to every consumer of CardScanLog.skip_reason — precisely the failure mode this test names. It stayed green. Three other tests in the file caught the mutation; this one could not.

What this changes

• Parametrised over _declared() — the values actually read out of the production modules. The same mutation now produces a red test_no_declared_value_is_empty_or_whitespace[ ].

• Added test_declared_roster_is_not_empty. The parametrised shape has a second way of going vacuous: if _declared() ever returns {} (moved directory, regex that stops matching, glob that finds nothing), pytest generates zero cases and the suite stays green while checking nothing. Verified by mutation — pointing CARDPICKER_DIR at a nonexistent path silently skips the parametrised test entirely and now trips this guard.

• Documented the subsumption. test_every_declared_skip_reason_value_is_unchanged is strictly subsumed by test_every_declared_constant_name_is_accounted_for (dict equality implies key-set equality). Kept, because its failure output is the two value sets — which names the changed production string directly, where the dict comparison buries it in a value-to-name-set diff. Its docstring now says it is a diagnostic and adds no mutation coverage of its own, so it stops being counted as a second check.

Mutation table

mutation	before	after
EXTRACTOR_FETCH_FAILED_SKIP_REASON = " "	🟢 green (3 other tests caught it)	🔴 red, by name
_declared() derivation broken (returns {})	🟢 green (0 cases generated)	🔴 red (test_declared_roster_is_not_empty)

Coordination with #584

PR #584 also touches this file. Its edits are in EXPECTED_SKIP_REASONS (lines 58-100) and appended after line 296; these are at lines 147-169. Disjoint, and this branch is based on origin/master, so either can merge first without conflict.

Verification

• test_skip_reason_roster.py — 42 passed (was 41).
• Full cardpicker/tests/ suite green.
• No production code changed.

🤖 Generated with Claude Code

https://claude.ai/code/session_013NhYmT1PxCcyemA16dFDxN