4.0 KiB
4.0 KiB
Working on tiny-wb
Structure and dependencies
- Keep the binary entry point small: configuration, diagnostics, terminal lifetime, and application startup belong there.
- Separate configuration/persistence, protocol decoding, transport, acquisition,
CSV export, shared scan state, and terminal presentation by responsibility.
Do not recreate a miscellaneous
structsmodule. - Protocol and export code must be usable without a terminal or physical device.
Accept
Read/Writeimplementations or explicit data instead of opening files or serial ports inside parsers. Keep model rules separate from transport rules. - Pass typed settings explicitly. Put user-adjustable defaults in
config/default.toml; keep wire-format invariants in the protocol module. Validate settings before connecting or creating output, including units, bounds, empty collections, and non-finite numbers. - Preserve existing configuration compatibility or document a tested migration. Unknown keys and invalid supplied values must produce actionable errors.
- Prefer small concrete types and functions over speculative abstraction. Introduce a trait only when multiple implementations or an actual test seam justify it. Avoid adding dependencies for small standard-library operations.
- Propagate I/O errors, use complete writes, bound receive buffers, and never
interpret binary payload as text. Do not use
unsafefor protocol ordering.
Documentation and comments
- Update user-visible behavior, configuration examples, architecture/source maps, protocol notes, and verification instructions in the same task as code changes. Link to the authoritative document instead of duplicating detailed guidance.
- Add module documentation explaining its responsibility and boundaries. Public APIs need Rustdoc covering units, invariants, errors, and meaningful examples. Private comments should explain intent, non-obvious constraints, or tradeoffs; do not narrate straightforward statements or retain commented-out code.
- Keep names and terminology consistent: MHz in user settings/CSV, Hz on the wire, dBm for levels, milliseconds for configurable delays and timeouts.
- Separate observed behavior, upstream protocol evidence, and unverified hardware assumptions. Preserve source links and verification dates for device facts. Never claim a simulated transcript is a hardware integration test.
- Document limitations candidly. Avoid vague TODOs; explain the missing behavior, its consequence, and what evidence would justify the next change.
- Keep README concise; use
docs/for operational and architectural details. Generated reports and badge SVGs stay ignored.
Verification
- Add focused regression tests for changed behavior, especially malformed input, boundary values, partial reads/writes, timeouts, absent ports, persistence, and failure cleanup. Use in-memory transcripts and isolated temporary directories.
- Run formatting, locked checks/tests, Clippy with warnings denied, documentation,
the release build, and the declared MSRV check. Use the exact commands in
docs/security-and-ci.md; run security gates when changing dependencies. - Run
cargo coverageafter changing Rust behavior. Raise the documented floor only from measured results; never exclude production code to satisfy a target. - Run
python3 scripts/test-badges.pyafter CI/badge changes. Verify documentation links and configuration examples after moving files or changing settings. - Report unavailable hardware, Jenkins, platform, or publication checks explicitly. Do not start Jenkins or publish shared assets as part of local verification.
Commits
- Preserve unrelated work. Make one commit per coherent task with a concise
imperative subject (
refactor:,feat:,fix:,test:,docs:, orci:). - Include tests and directly related documentation with their implementation. Keep an independent CI change or contributor policy in its own commit.
- Inspect each staged diff; never commit generated files, local configuration, logs, credentials, or build products. Do not push unless requested.