DevITWay | Pavel Volkov ac3a8a7c43 quality: MSRV, tarpaulin config, proptest for parsers (#84) ...

* fix: proxy dedup, multi-registry GC, TOCTOU and credential hygiene

- Deduplicate proxy_fetch/proxy_fetch_text into generic proxy_fetch_core
  with response extractor closure (removes ~50 lines of copy-paste)
- GC now scans all registry prefixes, not just docker/
- Add tracing::warn to fire-and-forget cache writes in docker proxy
- Mark S3 credentials as skip_serializing to prevent accidental leaks
- Remove TOCTOU race in LocalStorage get/delete (redundant exists check)

* chore: clean up root directory structure

- Move Dockerfile.astra and Dockerfile.redos to deploy/ (niche builds
  should not clutter the project root)
- Harden .gitignore to exclude session files, working notes, and
  internal review scripts

* refactor(metrics): replace 13 atomic fields with CounterMap

Per-registry download/upload counters were 13 individual AtomicU64
fields, each duplicated across new(), with_persistence(), save(),
record_download(), record_upload(), and get_registry_* (6 touch points
per counter). Adding a new registry required changes in 6+ places.

Now uses CounterMap (HashMap<String, AtomicU64>) for per-registry
counters. Adding a new registry = one entry in REGISTRIES const.
Added Go registry to REGISTRIES, gaining go metrics for free.

* quality: add MSRV, tarpaulin config, proptest for parsers

- Set rust-version = 1.75 in workspace Cargo.toml (MSRV policy)
- Add tarpaulin.toml: llvm engine, fail-under=25, json+html output
- Add coverage/ to .gitignore
- Update CI to use tarpaulin.toml instead of inline flags
- Add proptest dev-dependency and property tests:
  - validation.rs: 16 tests (never-panics + invariants for all 4 validators)
  - pypi.rs: 5 tests (extract_filename never-panics + format assertions)

* test: add unit tests for 14 modules, coverage 21% → 30%

Add 149 new tests across auth, backup, gc, metrics, mirror parsers,
docker (manifest detection, session cleanup, metadata serde),
docker_auth (token cache), maven, npm, pypi (normalize, rewrite, extract),
raw (content-type guessing), request_id, and s3 (URI encoding).

Update tarpaulin.toml: raise fail-under to 30, exclude UI/main from
coverage reporting as they require integration tests.

* bench: add criterion benchmarks for validation and manifest parsing

Add parsing benchmark suite with 14 benchmarks covering:
- Storage key, Docker name, digest, and reference validation
- Docker manifest media type detection (v2, OCI index, minimal, invalid)

Run with: cargo bench --package nora-registry --bench parsing

* test: add 48 integration tests via tower oneshot

Add integration tests for all HTTP handlers:
- health (3), raw (7), cargo (4), maven (4), request_id (2)
- pypi (5), npm (5), docker (12), auth (6)

Create test_helpers.rs with TestContext pattern.
Add tower and http-body-util dev-dependencies.
Update tarpaulin fail-under 30 to 40.

Coverage: 29.5% to 43.3% (2089/4825 lines)

* fix: clean clippy warnings in tests, fix flaky audit test

Add #[allow(clippy::unwrap_used)] to 18 test modules.
Fix 3 additional clippy lints: writeln_empty_string, needless_update,
unnecessary_get_then_check.
Fix flaky audit test: replace single sleep(50ms) with retry loop (max 1s).
Prefix unused token variable with underscore.

cargo clippy --all-targets = 0 warnings (was 245 errors)

2026-04-05 10:01:50 +03:00

8.9 KiB

Raw Blame History

View Raw