Testing

task check          # what CI runs: gofmt, vet, unit tests (race detector), build, script and privacy checks
task incus-test     # the real-systemd scenarios in containers (below)

go test ./...                              # unit tests, a few seconds

scripts/incus-test.sh                      # the real thing: Debian and Fedora under real systemd
scripts/incus-test.sh --distro arch        # also: arch (debian, fedora, arch)
scripts/incus-test.sh --only multirig,sites,direwolf
scripts/incus-test.sh --fresh              # start from brand new instances
scripts/incus-test.sh --cleanup            # delete the test instances

incus-test.sh needs incus. It builds hamstack, creates one system container per distribution (named hamstack-test-<distro>, and it never touches any other instance), and runs scripts/incus-scenarios.sh inside each as an unprivileged user with sudo, the way a real user would. The distributions run in parallel, each with a full log, and you get a summary.

The scenarios (about 190 assertions per distribution) cover: installing the programs; the guided setup; several rigs at once (their own ports, stopping, disabling, removing); sites and switching between them; migrating a config and units from the single-radio versions; Direwolf with a CAT radio and with a Digirig that is not plugged in; doctor (including the old-hamlib check); the group and linger handling across a new login and a restarted service manager; pat sync; a GPS (simulated receiver, a real gpsd); packet listen and call against a real Direwolf; the kernel AX.25 stack (only where the kernel has it: run it in a VM); a reboot (services must come back by themselves); and the interface driven through tmux. The first run downloads images and builds Mercury, so it takes several minutes; later runs reuse the instances and take about six.