Epic: Raise unit coverage toward critical-path targets

[mjkatgithub]

Planning

• Milestone: Phase 4
• Priority: P2
• Size: XL (split into sub-issues by domain)
• Estimate: 13+ SP (incremental; do not big-bang to 90%)

Goal

Increase meaningful unit coverage on critical paths and raise
vitest.config.ts thresholds gradually — without chasing 90% on pages/plugins
that are already covered by E2E.

Today: ~49% lines overall (npm run test:coverage); thresholds are
10–20% (intentionally low so CI passes). Target is ~90% on chosen
modules, not 90% repo-wide in one step.

Background

Low overall % is expected while many files have 0% coverage:

• app/pages/** (large, E2E-heavy)
• app/plugins/**
• Many app/utils/matrix*Permissions.ts stubs

High-value modules for unit depth first:

• app/composables/matrix/** and useMatrixClient.ts
• app/utils/chatTimeline/**
• Auth/signup composables and error mapping
• Unread/notification utils (Feature: Mention-priority unread notifications (@username) #35 area)

Follow-up to #122 and README "Test layers" section.

Scope (sub-issues to create)

• Coverage baseline report — document per-folder % in
  docs/testing-coverage.md; list top 10 uncovered critical files
• Matrix composables coverage — tests for messages.ts, rooms,
  session/crypto helpers; target ≥ 80% lines in app/composables/matrix/
• Chat timeline + unread utils — target ≥ 85% on
  app/utils/chatTimeline/ and core unread helpers
• Raise global thresholds (phase 1) — e.g. lines 30%, functions 40%,
  branches 35% after matrix + timeline slices land
• Raise global thresholds (phase 2) — step toward 50%/60% when pages
  get selective unit tests (not required for phase 1 merge)

Definition of Done

• Sub-issues created and linked under this epic
• Critical modules listed above meet their per-folder targets
• vitest.config.ts thresholds raised at least once with CI green
• README states repo-wide % vs module targets clearly

Nicht im Scope

• 90% on entire app/ in one milestone
• Coverage for generated/vendor code
• Lowering E2E count to compensate

Notes

• Use @vitest/coverage-v8 HTML report to find dead branches, not blind
  assertion padding.
• Exclude truly untestable glue from coverage.include only with justification
  in issue Notes.
• Each threshold bump should be its own small PR.

Related

• CI: Fix full lane (coverage + full E2E) #122 — full lane runs test:coverage on every MR
• CCD: max 1000 lines per test file; split specs when growing