dig-nat 0.4.0

Abstract NAT traversal for DIG Node peer connections — one connect() API over direct, UPnP/IGD, NAT-PMP, PCP, relay-coordinated hole-punch, and relay.dig.net as last-resort fallback; establishes an mTLS peer connection with peer_id = SHA256(TLS SPKI DER).
Documentation
name: CI

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

permissions:
  contents: read

concurrency:
  group: ci-${{ github.ref }}
  cancel-in-progress: true

env:
  CARGO_TERM_COLOR: always

jobs:
  test:
    name: Format / Clippy / Test / Docs
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: Install Rust toolchain
        uses: dtolnay/rust-toolchain@stable
        with:
          toolchain: stable
          components: rustfmt, clippy

      - name: Cache dependencies
        uses: Swatinem/rust-cache@v2

      - name: Check formatting
        run: cargo fmt --all --check

      - name: Check clippy
        run: cargo clippy --all-targets -- -D warnings

      - name: Build (release)
        run: cargo build --release

      - name: Check documentation
        run: cargo doc --no-deps

  coverage:
    name: Coverage (>=80% lines)
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: Install Rust toolchain
        uses: dtolnay/rust-toolchain@stable
        with:
          toolchain: stable
          components: llvm-tools-preview

      - name: Cache dependencies
        uses: Swatinem/rust-cache@v2

      - name: Install cargo-llvm-cov
        uses: taiki-e/install-action@cargo-llvm-cov

      - name: Install cargo-nextest
        uses: taiki-e/install-action@nextest

      # Unit + integration coverage, CI-gated at >=80% lines (ecosystem standard #157), run through
      # nextest (not a separate `cargo test`) so llvm-cov instruments the SAME nextest execution —
      # running nextest and llvm-cov as separate steps collects ZERO coverage and fails this gate
      # (#488). `--retries 2` retries a failing test up to 2x and reports it as flaky
      # (failed-then-passed) rather than a hard CI failure, surfacing flaky tests in the run output
      # (#489) while coverage collection stays intact.
      #
      # `method/upnp.rs`'s `RealIgd::add_port_mapping` is the ONE genuinely network-bound path — it
      # performs live UPnP SSDP multicast discovery + SOAP against a real Internet Gateway Device,
      # which no CI runner has. It sits behind the `IgdGateway` trait so the UPnP method's own logic
      # is unit-tested with a fake gateway (see tests/methods.rs), and its constructors are covered
      # (tests/peer_passthrough.rs); only the live SSDP call itself is unreachable in CI. Ignoring
      # this file keeps the gate stable + honest (same discipline as dig-relay ignoring its
      # OS-service syscall code). Every other module — including all NAT datagrams, STUN, the mTLS
      # dial, the relay client, and the mux transport — is exercised over loopback and counted.
      - name: Run tests with coverage (nextest, retries 2, gate >=80% lines)
        run: >-
          cargo llvm-cov nextest --all
          --retries 2
          --ignore-filename-regex 'method[\\/]upnp\.rs'
          --fail-under-lines 80
          --lcov --output-path lcov.info

      - name: Upload coverage report
        uses: actions/upload-artifact@v4
        with:
          name: lcov-coverage
          path: lcov.info
          if-no-files-found: error