Coverage: vite + frontend/providers (vite, inertia, broadcasting, storage, loader)

[tmgbedu]

Summary

Raises unit-test coverage for the Vite asset helper, the Loader utility and the inertia / broadcasting / storage service providers. Tests mock the filesystem, Vite manifest and request/network layers — no live servers or network required.

Coverage impact

Module	Before	After
vite/vite.py	24%	100%
loader/Loader.py	24%	100%
inertia/provider.py	50%	100%
storage/providers/provider.py	28%	100%
broadcasting/provider.py	19%	98%

Total project coverage: 73.94% (threshold fail_under = 68).

What's covered

• Vite — HMR (hot) vs manifest (production) modes, asset URLs (default / asset_url / custom resolver), tag + preload generation, CSP nonce, integrity, attribute resolvers, manifest caching & hashing, and the ViteException / ViteManifestNotFoundException error paths.
• Loader — get_modules, find / find_all, get_object(s), get_parameters, and the not-found / raise paths.
• Inertia provider — binding, middleware registration, and Jinja global injection (including the absent-templates branch and the page-markup helper).
• Storage provider — manager/driver registration and the /storage/{path} route (serve, 404, no-FastAPI early return).
• Broadcasting provider — manager/server bindings, stub publishing, the /broadcasting/auth handler across public/private channels and every user-resolution branch, routes/channels.py auto-loading, and the no-FastAPI fallbacks.

Notes

Two small pre-existing quirks were worked around in tests rather than changed (framework code left untouched): passing a plain dict to use_*_tag_attributes self-references the wrapping lambda, and the auth handler's Request annotation is an unresolved forward-ref under from __future__ import annotations, so its body is exercised directly.

Testing

uv run pytest --ignore=tests/masoniteorm/postgres — 1726 passed, 7 skipped.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[tmgbedu]

Code review verdict: REQUEST CHANGES (Ruff CI red) — content otherwise approve-worthy

(Formal request-changes not possible — same-account PR — so filing as a comment.)

Verified locally on task/vite-frontend-providers-coverage-721:

• Only 5 test files changed — no framework source touched. ✅
• Coverage matches claims exactly: vite 100%, loader 100%, inertia 100%, storage 100%, broadcasting 98% (uncovered: lines 77, 107), total 73.94% ≥ 68; 1726 passed, 7 skipped. ✅
• Tests assert real behavior, deps mocked, no live servers/network. ✅

Blocker — Ruff CI is failing (2 jobs). Reproduced locally:

1. ruff check F401: tests/broadcasting/test_provider.py:3 imports TestClient but never uses it — remove it.
2. ruff format --check reformats 3 files:
   • tests/broadcasting/test_provider.py (two over-wrapped await handler(...) calls)
   • tests/utils/test_loader.py (''' → """ on MODULE_SOURCE)
   • tests/vite/test_vite.py ("rel=\"modulepreload\"" → 'rel="modulepreload"')

Fix:

uv run ruff check --fix tests/broadcasting/test_provider.py
uv run ruff format tests/broadcasting/test_provider.py tests/utils/test_loader.py tests/vite/test_vite.py

Push and CI goes green — I'll re-approve immediately.

Follow-up bug tasks recommended (both correctly worked-around, not fixed — this is a test-only PR):

1. LOW — Vite.use_*_tag_attributes self-references its wrapping lambda when given a plain dict.
2. MEDIUM — broadcasting auth handler Request forward-ref (under from __future__ import annotations) can't be injected by FastAPI → 422 over real HTTP; endpoint effectively non-functional via normal request.

[tmgbedu]

✅ APPROVE (Code Reviewer)

Ruff blocker resolved in commit 51be36e2 (removed the unused TestClient import + ruff format on the 3 affected test files — test-only, no framework source touched, no stray files).

Final CI — all green:

• Ruff ✅ (both jobs)
• Pytest ✅ (3m14s)
• codecov/patch ✅

Confirmed locally too: ruff check . clean, ruff format --check . clean, 1726 passed, 7 skipped. Coverage unchanged — vite/loader/inertia/storage 100%, broadcasting 98%, total 73.94% ≥ 68.

Formal GitHub approval is blocked on a same-account PR, so recording approval here. Do not merge per standing order — over to the PM.

(Two follow-up framework bug tasks still recommended, unchanged: LOW Vite use_*_tag_attributes dict self-ref; MEDIUM broadcasting auth Request forward-ref → 422 over real HTTP.)