docs: project docs, scripts, deployment configs, and evidence

docs/evidence/ops/2026-03-27/quality/COVERAGE_REMEDIATION_20260327-233824.md

@@ -0,0 +1,100 @@
+# 2026-03-27 Q-004 Coverage Remediation Pass 5
+
+## Scope
+
+- Continue remediating the remaining real `Q-004` gaps after Pass 4.
+- This pass focused on:
+  - deeper non-network `internal/auth/providers` parsing and error branches
+  - low-coverage admin log pages
+
+## Changes
+
+### Frontend
+
+- Added page-level regression coverage in:
+  - `frontend/admin/src/pages/admin/LoginLogsPage/LoginLogsPage.test.tsx`
+  - `frontend/admin/src/pages/admin/OperationLogsPage/OperationLogsPage.test.tsx`
+- Newly covered page behavior includes:
+  - initial load
+  - filter application
+  - reset
+  - refresh
+  - pagination callback handling
+  - detail drawer open/close flow
+  - page-level error retry
+
+### Backend
+
+- Added provider non-network transport tests in:
+  - `internal/auth/providers/provider_http_roundtrip_test.go`
+- The new tests replace live network dependencies with a fake `http.DefaultTransport`.
+- Newly covered provider logic includes:
+  - `QQProvider.GetOpenID` success and invalid-JSON failure
+  - `QQProvider.GetUserInfo` API-error and success branches
+  - `WeiboProvider.ValidateToken` error / valid / ambiguous / invalid-JSON branches
+  - `WeChatProvider.ValidateToken` success / failure / invalid-JSON branches
+  - `GoogleProvider.ValidateToken` success and parse-failure branches
+  - `FacebookProvider.GetUserInfo` `error.message` and success branches
+
+## Verified Commands
+
+```powershell
+cd D:\project\frontend\admin
+npm.cmd run test:run -- src/pages/admin/LoginLogsPage/LoginLogsPage.test.tsx
+npm.cmd run test:run -- src/pages/admin/OperationLogsPage/OperationLogsPage.test.tsx
+npm.cmd run lint
+npm.cmd run build
+npm.cmd run test:coverage
+
+cd D:\project
+$env:GOCACHE='D:\project\.tmp\gocache'
+$env:GOPATH='D:\project\.tmp\go'
+$env:GOMODCACHE='D:\project\.tmp\go\pkg\mod'
+go test ./internal/auth/providers -run 'Test(QQProviderGetOpenIDAndUserInfoWithDefaultTransport|WeiboProviderValidateTokenWithDefaultTransport|WeChatProviderValidateTokenWithDefaultTransport|GoogleProviderValidateTokenWithDefaultTransport|FacebookProviderGetUserInfoWithDefaultTransport)$' -count=1
+go test ./... -count=1
+go vet ./...
+go build ./cmd/server
+go test ./internal/auth/providers ./internal/repository -cover -count=1
+```
+
+## Results
+
+### Frontend coverage
+
+- Overall:
+  - statements `56.81%`
+  - branches `44.67%`
+  - functions `57.38%`
+  - lines `57.57%`
+- Target areas:
+  - `services`: `86.2%` statements, `90.9%` branches
+  - `src/pages/admin/LoginLogsPage/LoginLogsPage.tsx`: `93.1%` statements, `55.55%` branches
+  - `src/pages/admin/OperationLogsPage/OperationLogsPage.tsx`: `91.52%` statements, `68.75%` branches
+
+### Backend coverage
+
+- `internal/auth/providers`: `28.7%`
+- `internal/repository`: `37.1%`
+
+## Validation Notes
+
+- `npm.cmd run test:coverage` completed successfully with `28` passing test files and `96` passing tests.
+- The same successful coverage run still printed one post-summary jsdom `AggregateError` network-noise line.
+  - This did not fail the command.
+  - It is still a real residual test-hygiene issue and should be traced separately instead of being ignored or misrepresented as a clean zero-noise run.
+
+## Real Conclusion
+
+- This pass materially reduced two real `Q-004` gaps:
+  - admin log pages are no longer among the main uncovered page-level hotspots
+  - `internal/auth/providers` improved substantially from the previous `15.2%`, but remains materially under-covered
+- `Q-004` still cannot be honestly declared closed.
+- The main remaining real gaps are now concentrated in:
+  - deeper `internal/auth/providers` paths beyond the new non-network parsing/error coverage
+  - still-uncovered admin pages/components such as:
+    - `PermissionsPage`
+    - `RolesPage`
+    - `ProfilePage`
+    - multiple admin drawers/modals that still remain near `0%`
+- Secondary remaining gap:
+  - `internal/repository` has improved, but `37.1%` is still not a closure-grade depth