pluriwave/openspec/changes/auto-media-art-quality/verify-report.md

Verification Report: auto-media-art-quality

Mode: Strict TDD (adversarial, fresh re-execution — apply-progress claims independently re-verified, not trusted) Date: 2026-07-19 Verdict: PASS WITH WARNINGS

Completeness

Source	Claim	Verified
tasks.md	22/22 tasks [x]	Confirmed by reading tasks.md directly — all checkboxes marked
apply-progress.md	22/22 tasks complete, single-batch delivery	Consistent with tasks.md and live code state

Test Execution (independent re-run, not trusted from apply report)

Command: flutter test test/servicios/navegacion_auto_test.dart --concurrency=1 --timeout=60s

Result: 25/25 passed (+0 through +24, All tests passed!). Matches apply-progress's claimed 25/25 exactly — re-run independently in this session, not copy-pasted from the report.

Spec Compliance Matrix

Spec Scenario	Covering Test(s)	Status
Valid, reachable remote logo	artUriPara: favicon https valido..., itemEmisora: usa el favicon remoto...	PASS
No logo (Case A: null/empty)	artUriPara: favicon null/vacio cae al drawable rotativo	PASS
Logo present but unreachable (Case B)	artUriPara: favicon malformado cae al mismo comportamiento rotativo	PASS for the malformed/non-http(s) subset only — see WARNING-1
Fallback art matches phone-UI per-station selection	indiceArtePara group, parity: phone/auto art order group	PASS — formula/order diff-verified against source, see below
Fallback art is on-brand, not launcher icon	Static: default_station_art.png deleted, 4 station_art_* drawables byte-identical to assets/images/ source	PASS
Root returns 3 folders, non-playable	ConstructorArbolAuto.raiz group (pre-existing, unchanged)	PASS
Empty folder returns empty list	ConstructorArbolAuto.hijos: lista vacia... (pre-existing)	PASS
Cold-start browse safety	FuenteEmisorasAutoLocal try/catch paths (pre-existing, unchanged)	PASS (not touched by this change)
Known codec+bitrate -> subtitle	subtituloCalidad matrix + itemEmisora: setea displaySubtitle...	PASS
Unknown codec/bitrate -> subtitle omitted, never literal "null"	subtituloCalidad matrix (incl. bitrate<=0, whitespace-codec) + null-substring test + itemEmisora: omite displaySubtitle...	PASS

Adversarial Checks (per orchestrator instructions)

1. Favicon validity gate — edge cases independently re-verified

Read lib/servicios/navegacion_auto.dart:31-42. Independently ran a throwaway Dart script (not trusting the apply report) to confirm the claimed Dart Uri quirk:

Uri.tryParse('http://') -> hasAuthority=true, host="", scheme=http

This confirms the apply-progress claim that a naive uri.hasAuthority check would wrongly accept 'http://'. The shipped code uses uri.host.isNotEmpty (not hasAuthority), which correctly rejects it — confirmed both by source inspection and by the passing faviconUsable test case ('http://': false).

Also independently re-ran Uri.tryParse against the other edge cases in the task list (cdn.example.com/logo.png, /relative/path/logo.png, not a url at all $$$ ///, ftp://cdn.example.com/logo.png) — all parse with an empty or non-http(s) scheme, so scheme == 'http' || scheme == 'https' alone already rejects them (the host.isNotEmpty check specifically fixes the 'http://' case). All 10 table-driven cases in the faviconUsable test pass.

Gap (see WARNING-1): the gate is a static URL-shape check only — it cannot detect a syntactically-valid https:// URL that 404s or points at a non-image response (spec's Case B literally lists "dead link, unreachable host, or non-image response" as triggers). This is a deliberate, documented design choice (design.md Decision 1), not an implementation bug — but it is a real narrowing of the spec's literal text.

2. Rotation algorithm parity — diffed, not assumed

Read both implementations side by side.

lib/widgets/tarjeta_emisora.dart:361-370 (_fallbackArtFor):

const arts = ['assets/images/station_art_aurora.png', '...cosmic...', '...pulse...', '...nova...'];
final index = seed.codeUnits.fold<int>(0, (a, b) => a + b) % arts.length;

lib/servicios/navegacion_auto.dart:21,47-48 (indiceArtePara + _nombresArte):

const _nombresArte = ['aurora', 'cosmic', 'pulse', 'nova'];
int indiceArtePara(String seed) => seed.codeUnits.fold<int>(0, (a, b) => a + b) % _nombresArte.length;

Confirmed identical: same fold expression, same modulus source (4 in both, since both lists have length 4), same element order (aurora, cosmic, pulse, nova). This is a provable, not assumed, match — both files were read and compared directly, not inferred from apply-progress prose.

3. Native drawables — byte-identical, dangling refs checked

cmp run on all 4 pairs — all reported IDENTICAL:

assets/images/station_art_aurora.png <-> android/.../drawable/station_art_aurora.png
assets/images/station_art_cosmic.png <-> android/.../drawable/station_art_cosmic.png
assets/images/station_art_pulse.png  <-> android/.../drawable/station_art_pulse.png
assets/images/station_art_nova.png   <-> android/.../drawable/station_art_nova.png

android/app/src/main/res/drawable/default_station_art.png confirmed absent (directory listing shows only the 4 new PNGs plus the pre-existing ic_stat_pluriwave.xml/launch_background.xml).

Grepped the whole repo for default_station_art — 8 hits, all in SDD artifact prose (openspec/changes/auto-media-art-quality/{apply-progress,tasks,design,proposal}.md and the prior android-auto-media change's artifacts). Zero hits in lib/ or live android/ resource/manifest files.

4. displaySubtitle fallback matrix — assertions inspected, not just names

Independently re-ran the test file and read the actual assertions (not just test names) in group('subtituloCalidad', ...):

Input	Assertion	Result
codec=mp3, bitrate=128	'128 kbps · MP3'	matches design table
codec=null, bitrate=128	'128 kbps'	matches
codec=mp3, bitrate=null	'MP3'	matches
codec=mp3, bitrate=0	'MP3' (bitrate<=0 -> unknown)	matches design's stated Radio Browser convention
codec=null, bitrate=null	isNull	matches — not '', not 'null'
codec=null, bitrate=0	isNull	matches
codec=' ' (whitespace), bitrate=null	isNull	matches — whitespace-only codec treated as unknown
codec=' mp3 ', bitrate=128	'128 kbps · MP3'	matches — trimmed correctly

Plus a dedicated test asserting result?.contains('null') ?? false is isFalse across 4 representative cases — directly defends the "never render literal null" requirement rather than relying only on exact-match equality. All assertions call real production code (subtituloCalidad); none are tautological, none are ghost loops.

5. AI attribution / debug prints / dead code

• git diff on both changed files scanned for print(, debugPrint, console.log, TODO, FIXME, Anthropic, Claude, Co-Authored, Generated by — zero matches.
• Grepped lib/ for _defaultArtUri and _artUriPara (the two retired identifiers) — zero matches anywhere, confirming full removal, not just from the one call site.

6. Working tree state

git status --porcelain output:

 D android/app/src/main/res/drawable/default_station_art.png
 M lib/servicios/navegacion_auto.dart
 M test/servicios/navegacion_auto_test.dart
?? android/app/src/main/res/drawable/station_art_aurora.png
?? android/app/src/main/res/drawable/station_art_cosmic.png
?? android/app/src/main/res/drawable/station_art_nova.png
?? android/app/src/main/res/drawable/station_art_pulse.png
?? openspec/changes/auto-media-art-quality/

Confirmed: unstaged modifications/deletion plus untracked new files only. No commit was made by the apply agent, matching its claim.

7. Test count cross-check

Re-ran navegacion_auto_test.dart independently in this session (see "Test Execution" above): 25/25 passed, matching apply-progress's claimed 25/25 exactly. Not trusted from the report — executed fresh in this session.

TDD Compliance

Check	Result	Details
TDD Evidence reported	Yes	"TDD Cycle Evidence" table present in apply-progress.md, one row per phase
All tasks have tests	Yes	Every behavioral task (1.1-6.2) has a corresponding test group in navegacion_auto_test.dart; Phase 7 tasks correctly marked static-review-only
RED confirmed (tests exist)	Yes	All claimed test groups exist in the live test file
GREEN confirmed (tests pass)	Yes	25/25 passing on independent re-run
Triangulation adequate	Yes	Each behavior has 2+ distinct test cases with varying expected values
Safety Net for modified files	Yes	navegacion_auto.dart modified — full test file (25 tests, incl. 8 pre-existing) re-run and green, no regression

TDD Compliance: 6/6 checks passed

Assertion Quality

File	Line(s)	Assertion	Issue	Severity
test/servicios/navegacion_auto_test.dart	120-129	if (indiceArtePara(a.uuid) != indiceArtePara(b.uuid)) { expect(...) }	Conditional/guarded assertion — could execute zero assertions if the two literal seeds hashed to the same index	SUGGESTION

Independently computed indiceArtePara('uuid-a') == 1 and indiceArtePara('uuid-bbbb') == 0 — confirmed the guard condition is true for the literals actually used today, so the assertion does execute. Minor fragility only (a future seed-literal edit could silently disable the check); the same property is fully and unconditionally enumerated by the parity: phone/auto art order test (Phase 6.2), so coverage is not actually at risk.

No tautologies, no ghost loops over possibly-empty collections, no assertion-free tests found.

Assertion quality: 0 CRITICAL, 0 WARNING, 1 SUGGESTION

Design Coherence

Design Decision	Code Match
Static favicon validity gate (no network), malformed-subset-only	Matches exactly — faviconUsable is a pure sync function, no I/O
Port _fallbackArtFor verbatim (formula + order)	Confirmed identical via direct source diff
Retire default_station_art.png + _defaultArtUri (rotation is total)	Confirmed — both removed, zero dangling references
displaySubtitle format, omit when both unknown	Matches design's table exactly, incl. bitrate<=0 and whitespace-codec edges
_nombresArte stays private; order verified indirectly via artUriPara	Documented deviation in apply-progress, judged reasonable

No design deviations that break a spec requirement.

Issues Found

CRITICAL

None.

WARNING

WARNING-1 — Case B coverage is narrower than the spec's literal text (documented/accepted, flagged for archive sign-off)

Spec text (specs/android-auto-media/spec.md, "Station's logo URL is present but unreachable (Case B)") lists the triggering conditions as "dead link, unreachable host, or non-image response." The shipped faviconUsable gate is a static URL-shape validator only — a syntactically valid https://cdn.example.com/dead-link.png that returns 404, or a URL resolving to a non-image response, still passes the gate and is handed to the OS art loader as-is, which may still render a broken/blank tile. This contradicts the spec's stated "the car head unit MUST NOT display an empty, broken-image, or indefinitely-loading art tile for that station" for those specific sub-cases.

This is not an implementation bug — design.md Decision 1 explicitly considered and rejected live-reachability checking (HTTP HEAD, TTL cache, etc.) due to TOCTOU risk and the untested handler-path constraint, and states the proposal's "Split Judgment explicitly permits this narrowing of A." The gap is real and testably confirmed (only the malformed/non-http(s) subset is covered; genuine live-404 or non-image-response cases cannot be tested without network I/O, and are not tested). Recommend explicit archive-time sign-off that this narrowing is accepted, since the literal spec scenario text is broader than what ships.

SUGGESTION

SUGGESTION-1 — Guarded/conditional assertion in artUriPara "uuids distintos" test

See Assertion Quality table above. Low priority; the same guarantee is already fully covered by the unconditional parity: phone/auto art order test.

Next Steps

• sdd-archive is appropriate — no CRITICAL issues block archival. Recommend the archive step (or the user) explicitly acknowledge WARNING-1 (spec's Case B literal scope vs. shipped static-only gate) as an accepted, documented limitation before closing the change.
• Phase 7/8's static-review-only items (native drawable resolution via android.resource://, real Android build/DHU verification) remain unverified by this report for the same reason apply-progress could not verify them: no Android build environment available in this session. A real device/DHU pass is still recommended before shipping, per both apply-progress and design.md's own open question.