Skip to content

Tests: medir y visualizar cobertura del frontend #2962

Closed
Closed
Tests: medir y visualizar cobertura del frontend#2962
Labels
enhancementNew feature or request

Description

@mblua
mblua
opened on Oct 8, 2026

Objetivo

Conocer qué proporción del código fuente del frontend TypeScript/Solid ejecutan nuestros tests y localizar los archivos y caminos sin probar.

Contexto verificado

En main, SHA 400a351, package.json ejecuta Vitest mediante npm test. vitest.config.ts selecciona tests en src, pero no configura cobertura; no se encontró generación de reportes de cobertura en los workflows revisados.

Alcance y aceptación

  • Obtener una medición reproducible con la suite existente, indicando commit, comando y resultado de los tests.
  • Generar un reporte HTML navegable por archivo con cobertura de líneas, funciones y ramas, indicando las métricas que soporte la herramienta elegida.
  • Incluir el código fuente elegible aunque ningún test lo importe; explicitar las exclusiones y su motivo, sin contar tests o código generado como producto.
  • Entregar un resumen global y por archivo, con las zonas sin cobertura y las limitaciones de la medición.
  • Registrar instrucciones mínimas para repetir la medición y ubicar el reporte.

Límites

Issue de seguimiento solicitado por el usuario el 2026-10-08. Pendiente de implementación. La elección técnica se confirmará con los especialistas al retomarlo. No incluye aumentar cobertura, agregar tests, fijar porcentajes mínimos, bloquear PRs ni publicar reportes en servicios externos o CI; esas decisiones quedan para evaluar después del diagnóstico.

La cobertura mide ejecución; no demuestra que las aserciones validen correctamente el comportamiento.

Activity

  1. mblua
    added
    enhancementNew feature or request
    on Oct 8, 2026

  2. mblua commented on Oct 10, 2026

    @mblua
    mblua
    on Oct 10, 2026
    OwnerAuthor

    Implementation requested by the user on 2026-10-10: measure and visualize frontend coverage using the existing suite. Architect selected the existing Vitest runner with exact @vitest/coverage-v8 4.1.5, lockfile and opt-in coverage configuration. Allowed tracked files: package.json, package-lock.json, vitest.config.ts.

    Technical scoring (task-technical-scoring):

    Category Score Anchor/evidence
    Scope footprint 7 Three small configuration/dependency files, below coherent-module anchor 10.
    Diagnosis difficulty 2 Missing provider/config verified; near stated-cause anchor 1.
    Domain knowledge 6 Coverage/remapping rules, below subsystem anchor 10.
    Solution-space ambiguity 3 Architect chose single existing runner over external runners; near determined-change anchor 1.
    Blast radius and reversibility 3 Opt-in local config, reversible dependency/config changes; near local anchor 1.
    Dependency and integration load 9 One external version-matched provider and lock; near external-library anchor 10.
    Verification difficulty 8 Must compare eligible sources with report and prove non-imported runtime source appears at zero; focused positive control and remapping inspection.
    State, data and compatibility 1 No persisted application state or schema.
    Concurrency and timing 1 No new concurrent application behavior.
    Environment and tooling 7 Node/Vitest/Vite versions verified; coverage-provider installation/remapping detail below platform/toolchain anchor 10.

    Raw sum: 47. Final: round((47 - 10) / 190 * 99) + 1 = 20/100.

    Excluded ceremony: issue/branch/PR operations, review rounds, delegation, CI wait, merge/cleanup and approval handling.

    Delivery: Full, one unit expected (architect confirms partition rule). Express and Lite are ineligible due to the new dependency. Architect authors; owning UI developer and Grinch independently review; all three name the same Plan-SHA256. Verification difficulty >=8 requires Grinch review of the positive-control proof. No other category veto applies. Human approval of the exact reviewed plan is required before implementation. No new tests, thresholds, CI/report publication, scripts or application product changes.

  3. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    User explicitly approved on 2026-10-15 the architect's proposed scope extension: src/sidebar/stores/project.ts, src/sidebar/stores/project.test.ts, src/sidebar/components/ProjectPanel.loop-automation.test.tsx. Preserve the AST boundary test and evaluate complete coverage on Node22 with one worker. This authorizes scope; a new immutable plan and exact-version consensus precede implementation.

    Evidence: room-shared/2962-architect-blocker-proposal-20261012.md, 2962-two-blockers-diagnosis-20261012.md and 2962-focused-20261012T1515Z/report.md. Existing local commit 1ef02b5 changes only provider/lock/config; not published. Focused Node22 tests pass but ws rejection persists without coverage. Exact historical caller remains unproven.

    Remaining-delta scoring (previous score raw47/final20 retained in earlier comment):

    Category Score Anchor/evidence
    Scope footprint 6 Three additional existing source/test files; below coherent-module anchor10.
    Diagnosis difficulty 8 Concrete stale-continuation path and runtime errors; below unfamiliar-subsystem anchor10, live caller not fully proven.
    Domain knowledge 5 Solid store lifecycle, deferred promises and test isolation; between ordinary code1 and subsystem10.
    Solution-space ambiguity 5 Architect selected scoped existing guards/test isolation; some execution uncertainty, below alternatives anchor10.
    Blast radius and reversibility 5 Shared snapshot validity path, reversible localized guard; below shared-path anchor10.
    Dependency and integration load 3 Existing APIs/fake transport only; near self-contained1.
    Verification difficulty 10 Existing deferred harness supports before/after discriminating regression; source effect plus complete coverage acceptance required.
    State, data and compatibility 3 Ephemeral generation/request state only; no persisted migration.
    Concurrency and timing 11 Late promise resolution/rejection after invalidation and cleanup; near async ownership anchor10.
    Environment and tooling 6 Verified isolated Node22; existing runner and one-worker coverage; below toolchain anchor10.

    Raw62. Final=round((62-10)/190*99)+1=28/100. Excluded ceremony: planning/review rounds, issue/branch/PR, delegation, CI wait, merge/cleanup and approvals.
    Continue Full, architect-authored contract; concurrency11 requires architect and verification10 requires Grinch proof review. Same initial participants architect/UI/Grinch approve the exact new digest. No implementation until applicable renewed step7.5. No new dependencies, AST test edits, browser resolver/socket mocks, error suppression or unrelated fixes. Rust I2963 scope unresolved.

  4. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    I2962 execution exposed a plan discrepancy, not new product scope: v5 requires a new snapshot after the queued follow-up reload deliberately fails before requesting snapshots. Stop state and evidence preserved in room-shared/2962-ui-v5-20261015T1610Z/report.md. Before: all12 stale getter regressions fail discriminantly; after: store48/boundary26/UI7 pass, one fixture wait fails. No unhandled ws reported in this focused run; full-suite acceptance not established.

    Remaining delta rescoring (earlier scores raw47/final20 and raw62/final28 retained):

    Category Score Anchor/evidence
    Scope footprint 2 Local contract/expectation in existing test file, near small-file anchor1.
    Diagnosis difficulty 2 Failure path and impossible new-snapshot wait identified directly, near stated cause1.
    Domain knowledge 3 Existing deferred reload/snapshot lifecycle, below subsystem10.
    Solution-space ambiguity 2 Local failure-path wait contract, near determined-change1; architect decides exact shape.
    Blast radius and reversibility 2 Existing local test only, reversible, near local1.
    Dependency and integration load 1 Existing fixture/runner; self-contained.
    Verification difficulty 5 Existing targeted check plus preserved discriminating evidence; below new-test harness10.
    State, data and compatibility 1 No persisted-state change.
    Concurrency and timing 5 Awaited deferred failure path and cleanup; below async ownership10.
    Environment and tooling 2 Shared verified Node22/npm11.6.2 already available, near identical environment1.

    Raw25. Final=round((25-10)/190*99)+1=9/100. Ceremony excluded: author/reviewer rounds, messaging, purge, issue/branch/PR/CI wait/merge.
    The cumulative delivery remains Full with async product guard/Verification10 evidence and initial architect/UI/Grinch participants. New immutable complete version must reconcile this wait contract without broadening files, modifying product flow, raising timeouts or suppressing errors. No repair continuation before applicable consensus and step7.5. Existing dirty three repair paths must be preserved; no reset/cleanup. All other valid evidence is reusable.

  5. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    User explicitly approved the architect's four-file scope extension: three existing source guards switch eager raw acquisition to lazy glob keys plus direct UTF8 reads, preserving exact source universe/text/order/assertions/parser/timeouts; vitest.config.ts excludes only the proven test-only src/sidebar/components/settings/agentReorderDnd.testkit.ts. No source helper is edited. Nine cumulative modified paths, no new/deleted files.

    Confirmed diagnosis: room-shared/2962-ui-acceptance-20261015T1645Z/report.md and 2962-confirmed-minimal-scope-20261015.md. Diagnostic real main has nonempty zero-hit mappings; raw canonical same-path empty mapping suppresses untested processing. Original successful 3846-test run remains historical; its coverage percentages are not accepted product metrics.

    Remaining-delta technical scoring (previous raw47/final20, raw62/final28, raw25/final9 retained):

    Category Score Anchor/evidence
    Scope footprint 6 Four localized existing test/config files; below coherent module10.
    Diagnosis difficulty 3 Concrete collision already discriminated, near stated-cause1.
    Domain knowledge 5 Vite raw imports and coverage-map identity, below subsystem10.
    Solution-space ambiguity 2 Architect chose existing keys/fs pattern, near determined-change1.
    Blast radius and reversibility 4 Guards must preserve source universe/text; local reversible test/config change below shared-path10.
    Dependency and integration load 2 Existing fs shim, glob pattern and runner; near self-contained1.
    Verification difficulty 10 Must prove same inputs/asserts and real unimported zero control in measured denominator; existing harness/reporting supports it.
    State, data and compatibility 1 No persisted application-state migration.
    Concurrency and timing 1 No new concurrent application behavior.
    Environment and tooling 6 Existing verified toolchain, real/raw remap and full-run output below platform/toolchain10.

    Raw40. Final=round((40-10)/190*99)+1=17/100. Ceremony excluded: issue/branch/PR operations, consensus/review rounds, messaging/purge, waiting for CI and merge/cleanup.
    Cumulative delivery remains Full with the same initial architect/UI/Grinch participants. Verification10 requires independent proof review. New immutable complete version and renewed exact-version consensus/step7.5 precede implementation. No manual coverage JSON fixes, importing main artificially, broad exclusions, new fixtures/custom renderer/provider, timeout changes or jsdom repairs. Scope approval does not change Rust I2963.

  6. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    PR3025 remains blocked by test-debt on exact head bd7384f. The one authorized same-job rerun completed with the identical three new jscpd clones in ProjectPanel.loop-automation.test.tsx. Both logs are preserved in room-shared/2962-ci-38778703194/test-debt-attempt1.log and test-debt-attempt2.log; no further retry is planned without a code fix.

    New clone pairs:298–315 vs574–591;622–640 vs742–760;655–699 vs768–812. The changed test belongs to the already authorized I2962 regression/fixture scope. Required correction must preserve the full case matrix, assertions, deferred sequencing, snapshot identity/count checks and cleanup while satisfying the existing duplication convention. No baseline/threshold/ignore/suppression changes, no extra files or product behavior. Architect owns the exact helper shape under docs/testing/test-code-duplication.md.

    Remaining-delta score (earlier scores retained):

    Category Score Anchor/evidence
    Scope footprint 2 One existing test file/local blocks, near small-file1.
    Diagnosis difficulty 2 Exact repeated ranges reproduced twice, near stated cause1.
    Domain knowledge 3 Existing UI/deferred test pattern, below subsystem10.
    Solution-space ambiguity 3 Existing duplication convention constrains choice; architect resolves exact extraction.
    Blast radius and reversibility 3 Local test-only restructuring, reversible, below shared-path10.
    Dependency and integration load 1 No new dependencies, self-contained.
    Verification difficulty 5 Existing duplication gate and existing regression tests; below new-harness10.
    State, data and compatibility 1 No persisted state or migration.
    Concurrency and timing 2 Preserve existing deferred sequencing, no new concurrency.
    Environment and tooling 2 Shared verified Node22/npm11.6.2 and installed detector already available.

    Raw24. Final=round((24-10)/190*99)+1=8/100. Ceremony excluded: round/purge/review/issue/branch/PR/CI wait/merge operations. Cumulative workflow remains Full and same initial architect/UI/Grinch consensus with exact digest. A complete immutable next version must reconcile the test-only correction before implementation, reusing unchanged coverage/runtime/source proof where valid; changed-head coverage and CI will still be required. Prior main delta2c18a98d was independently classified unrelated; no automatic merge/update just for the count of commits behind.

  7. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    I2962 — remaining delta after executing v11

    The exact three v11 extractions were implemented in the sole authorized test, without a commit. UI 8/8, guards 28/28, typecheck, frontend-dependencies, version and whitespace checks passed. dup:changed against actual origin/main failed with one genuine residual clone between the extracted delete helper and the existing cancel case (14 lines/102 tokens), plus three clones from earlier ignored backup files. Current test SHA256: 36f7b123307e537eb5f775bcc3c952a79ad261c49589d61fbc73f124818ce8d9. HEAD remains bd7384f; origin/main remains 2c18a98. No commit or push; CI is not green.

    Remaining-delta technical score, using task-technical-scoring:

    Category Score Anchor and evidence
    Scope footprint 1 One existing test, one residual repeated block; no additional tracked files.
    Diagnosis difficulty 1 Detector identifies exact matching spans and backup paths.
    Domain knowledge 2 Familiar TypeScript and existing DOM test helpers.
    Solution-space ambiguity 3 Architect must close the local extraction contract while retaining distinct actions.
    Blast radius and reversibility 2 Local test-only edit; preserve artifacts and source recovery bytes.
    Dependency and integration load 1 No external dependencies or new shared modules.
    Verification difficulty 3 Existing focused test and duplication gate directly exercise acceptance.
    State, data and compatibility 1 No persisted or serialized state changes.
    Concurrency and timing 2 Preserve existing async waits; discovery sequence already passes.
    Environment and tooling 3 Ignored backup test sources enter the shipped detector; preserve them without changing detector/config.

    Raw 19; final round((19-10)/190*99)+1 = 6/100. Prior remaining delta raw24/final8. Reduced remaining work does not lower the accumulated Full workflow, original architect/UI/Grinch consensus contract or Verification10 proof requirement. Excluded ceremony: issue/branch/PR work, planning and review rounds, messaging, waiting for CI, merge and cleanup.

    Execution evidence: room-shared/2962-v11-ui-blocker-20261015T2037Z/report.md, dup.log, implementation.patch and before/after hashes. v11 consensus succeeded; execution showed a plan correction is necessary. Next round must preserve the current uncommitted test and evidence, avoid baseline/ignore/threshold edits, and keep original assertions/product behavior. No unrelated issue or scope expansion is authorized.

  8. mblua
    added a commit that references this issue on Oct 15, 2026

    test: reuse loop automation interactions (#2962)

    9f819d3

  9. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    The remaining duplication blocker is resolved in reviewed commit 9f819d3f1e90ff19cf924ca1b35ab83e9a5e78a5, now published in PR #3025. Actual duplication gate: 270 files, 117 existing clones, zero new clones. Task-owned backups were preserved outside the scan with matching file sets, hashes and sizes; no detector/configuration/baseline change was made.

    Fresh complete coverage on that clean commit passed all 270 test files / 3,846 tests:

    • Lines: 18,442 / 20,679 = 89.18%.
    • Functions: 6,688 / 7,568 = 88.37%.
    • Branches: 9,551 / 11,945 = 79.95%.
    • Statements: 23,223 / 26,699 = 86.98%.

    Typecheck/build and applicable local guards passed; independent code review, Verification10 and UI universe/closure/static-HTML proof passed. The 182 eligible source files match JSON/summary exactly; unimported runtime main.tsx remains included at zero hits. The known range-end limitation affects some HTML highlight extents, with counters and reported metrics validated independently.

    Remote CI for this exact new head is pending; no failed check is waived. Keep #2962 open until protected PR landing and fresh merged-main measurement/proof are recorded. No final app build or installation was requested.

  10. mblua
    added a commit that references this issue on Oct 15, 2026

    Merge pull request #3025 from mblua/feature/2962-frontend-coverage

    a1a694b

  11. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    PR #3025 is merged at a1a694b5962f30f265c1a07b5b91eace18fb683a (2026-10-15T22:15:57Z), after final verification of all applicable exact-head CI on 9f819d3f1e90ff19cf924ca1b35ab83e9a5e78a5. The previously failing test-debt gate now succeeds. Official full-tier jobs were skipped by the existing full=false workflow condition; no failing check was bypassed.

    The room checkout is clean on main, equal to origin/main and its upstream, and feature-branch cleanup is complete. A fresh complete coverage measurement is assigned on this exact merged-main SHA using the same pinned shared runtime and a new output directory. Keep this issue open until that final measurement and applicable proof are accepted; the earlier 89.18% line coverage remains the validated feature-head measurement meanwhile.

  12. mblua commented on Oct 15, 2026

    @mblua
    mblua
    on Oct 15, 2026
    OwnerAuthor

    Delivered through PR #3025, merged as a1a694b5962f30f265c1a07b5b91eace18fb683a on 2026-10-15T22:15:57Z. All applicable CI passed on the exact reviewed feature head; official full-tier skips follow the existing full=false workflow condition. The duplication blocker is resolved without changing the detector or baseline.

    Fresh final measurement on clean merged main passed all 270 test files / 3,846 tests using Node22.23.3, npm11.6.2 and the pinned Vitest/V8 provider4.1.5:

    Metric Covered / total Final coverage
    Lines 18,442 / 20,679 89.18%
    Functions 6,688 / 7,568 88.37%
    Branches 9,551 / 11,945 79.95%
    Statements 23,223 / 26,699 86.98%

    The inventory contains 466 tracked TS/TSX files, 284 explicitly excluded test/helper/declaration sources and 182 eligible files. JSON and summary contain exactly that eligible universe. Unimported runtime main.tsx remains included with nonempty executable mappings at zero hits; type-only0/0 rows are not runtime omissions.

    Independent final UI proof and Grinch Verification10 both passed on this exact merged SHA. They rechecked report integrity, totals, the zero-execution control, input/closure equivalence and static HTML navigation. All 57,150 checked local HTML links/resources/anchors resolve. The known 1,049 inverted range ends across 53 files can make highlight extents imprecise; validated counters and the four reported metrics are unaffected. This remains Node/jsdom execution coverage, not browser/Tauri correctness or exhaustive token mapping fidelity.

    Final local HTML: artifacts/coverage/frontend/a1a694b5962f30f265c1a07b5b91eace18fb683a-20261015T221802Z/index.html. Complete measurement receipt: room-shared/2962-execution-20261015T221802Z/report.md. Final proofs: room-shared/2962-final-main-ui-proof-20261015T222727Z/report.md and room-shared/2962-final-main-grinch-verification10-20261015/review.md. Original JSON/summary/HTML and hash/byte manifests are preserved without edits or external upload.

    Reproduce from the repo with the same Node/npm versions and installed lockfile dependencies, using a fresh output directory:

    export COVERAGE_OUTPUT="artifacts/coverage/frontend/$(git rev-parse HEAD)-$(date -u +%Y%m%dT%H%M%SZ)"
npm test -- --coverage --maxWorkers=1

    The room checkout is clean on main, equal to origin/main and its upstream; feature cleanup is complete. Frontend build validation passed; no final application build or installation was requested. All acceptance gates for #2962 are complete. Rust coverage remains separately tracked in #2963 with its scope decision pending.

  13. mblua
    closed this as completedon Oct 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions