fix(scanner): read verifies markers from shell scripts (REQ-319, #870)

[avrabe]

Closes #870.

coverage --tests read # rivet: verifies REQ-X from a Python file but not
the identical comment from a shell script. detect_language had no entry for
sh at all, so scan_file returned before any pattern ran — the file was
skipped in silence.

a.rs   // rivet: verifies REQ-RS-001   →  found
b.py   #  rivet: verifies REQ-PY-001   →  found
c.sh   #  rivet: verifies REQ-SH-001   →  No test markers found

This costs evidence, not tidiness

Not every falsifiable check is a unit test, and the ones that aren't tend to
guard the nastiest failures. Your two examples — a gate refusing any workflow
that redirects key material into a file (proved by rejecting four shapes,
including the exact line your own deposit workflow shipped), and one asserting
the installer warns when a different binary wins the PATH lookup — both run in
CI, both carry negative controls, and both are the sole evidence for their
requirement. Which therefore could not reach verified by any honest route.

The fix

Shell gets its own language category rather than an alias for python: the
comment pattern is identical but the enclosing-function regex is not
(name() { and function name, not def).

Also a shebang fallback for extensionless files, since a gate is frequently
tools/no-key-on-disk rather than …​.sh. Consulted only when a file has no
extension, so a tree of images and binaries isn't read looking for one — and the
test pins that a plain text file with no shebang is still skipped.

Both halves negative-controlled:

remove the sh mapping       →  extension test FAILED
remove the shebang fallback →  extensionless test FAILED

On #871 — part (1) is already fixed

Part (1) shipped in v0.34.0 (REQ-294). Verified on current main:

swe3-refines-swe2   sw-detail-design   0   0   n/a%
V-closure: sw-detail-design (all 2 rules)   n/a%  [0/0]
JSON: percentage: null, empty_scope: true

That's exactly the shape your suggestion (1) asks for, including the JSON null.
You're on scry's vendored copy, which predates the release — worth a bump.

Part (2) is real and unaddressed and is now REQ-320: there's no way to subset
an embedded preset's rules, so adopting aspice for SWE.1/SWE.6 necessarily
inherits SWE.2/3/4 rows you can never satisfy. Your workaround — a checked-in
declaration of unmodelled levels, gated in CI, failing both when an empty
population is undeclared and when a declaration goes stale — is the tell that
this is a mechanism rivet should own.

I've deliberately grouped it with REQ-309 (exempt-when-field) and REQ-313
(declared-blocked criteria) rather than solving it standalone: three
near-identical mechanisms for "this is declared, not forgotten" would be worse
than one.

Gates

cargo fmt --check · clippy --all-targets -D warnings on 1.97.0 ·
cargo test --workspace (exit 0, 65 ok, 0 failed) · rivet validate ·
rivet docs check — all exit 0. REQ-319 → implemented, 2 test markers.

[codecov]

Codecov Report

❌ Patch coverage is 87.83784% with 9 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
rivet-core/src/test_scanner.rs	87.83%	9 Missing ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

📐 Rivet artifact delta

Change	Count
Added	2
Removed	0
Modified	0
Downstream impacted (depth ≤ 5)	0

Graph

graph LR
  REQ_319["REQ-319"]:::added
  REQ_320["REQ-320"]:::added
  classDef added fill:#d4edda,stroke:#28a745,color:#155724
  classDef removed fill:#f8d7da,stroke:#dc3545,color:#721c24
  classDef modified fill:#fff3cd,stroke:#ffc107,color:#856404
  classDef overflow fill:#e2e3e5,stroke:#6c757d,color:#495057,stroke-dasharray: 3 3

Loading

Added

• REQ-319
• REQ-320

📎 Full HTML dashboard attached as workflow artifact rivet-delta-pr-872 — download from the workflow run.

_{Posted by rivet-delta workflow. The graph shows only changed artifacts; open the HTML dashboard (above) for full context.}

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rivet Criterion Benchmarks'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: b3cc8a4	Previous: 9d9a697	Ratio
store_insert/100	83438 ns/iter (± 2949)	66482 ns/iter (± 3182)	1.26
store_insert/1000	890745 ns/iter (± 9791)	718522 ns/iter (± 3126)	1.24
store_lookup/100	2068 ns/iter (± 32)	1461 ns/iter (± 4)	1.42
store_lookup/1000	26434 ns/iter (± 107)	17816 ns/iter (± 33)	1.48
store_lookup/10000	355702 ns/iter (± 1284)	258802 ns/iter (± 2208)	1.37
store_by_type/100	95 ns/iter (± 0)	75 ns/iter (± 1)	1.27
store_by_type/1000	95 ns/iter (± 0)	75 ns/iter (± 0)	1.27
store_by_type/10000	94 ns/iter (± 0)	75 ns/iter (± 0)	1.25
schema_load_and_merge	1511704 ns/iter (± 28167)	1183968 ns/iter (± 12458)	1.28
link_graph_build/100	157920 ns/iter (± 842)	126045 ns/iter (± 264)	1.25
link_graph_build/1000	1902548 ns/iter (± 16006)	1497016 ns/iter (± 23213)	1.27
link_graph_build/10000	27180072 ns/iter (± 1386691)	21409503 ns/iter (± 500407)	1.27
validate/100	504193 ns/iter (± 3302)	344980 ns/iter (± 1286)	1.46
validate/1000	18318731 ns/iter (± 142208)	10667637 ns/iter (± 38321)	1.72
validate/10000	1521291383 ns/iter (± 11816313)	753996489 ns/iter (± 5648610)	2.02
traceability_matrix/100	4355 ns/iter (± 70)	3209 ns/iter (± 30)	1.36
traceability_matrix/1000	59664 ns/iter (± 475)	33994 ns/iter (± 115)	1.76
traceability_matrix/10000	821192 ns/iter (± 3415)	554924 ns/iter (± 2007)	1.48
diff/100	62556 ns/iter (± 370)	47962 ns/iter (± 113)	1.30
diff/1000	712804 ns/iter (± 3944)	516450 ns/iter (± 1949)	1.38
diff/10000	7730356 ns/iter (± 152272)	6242821 ns/iter (± 139231)	1.24
query/100	1114 ns/iter (± 11)	765 ns/iter (± 3)	1.46
query/1000	14044 ns/iter (± 311)	10410 ns/iter (± 33)	1.35
query/10000	318601 ns/iter (± 5604)	170185 ns/iter (± 625)	1.87
document_parse/10	23671 ns/iter (± 333)	16350 ns/iter (± 41)	1.45
document_parse/100	170454 ns/iter (± 984)	111521 ns/iter (± 478)	1.53
document_parse/1000	1565569 ns/iter (± 30845)	1038384 ns/iter (± 8076)	1.51

This comment was automatically generated by workflow using github-action-benchmark.