tsunagi/docs/testing.md

Testing

How to run everything is in ../README.md. Rules for writing tests are in ../AGENTS.md.

Ground rules

Tests use real iroh endpoints on loopback, real handshakes, real SQLite in per-test temporary directories, and independent agent instances. Discovery is substitutable; iroh, authentication, message passing and persistent storage are not.

The suite runs with no internet, no public DHT, no public relay, no administrator rights and no changes to OS network settings: endpoints bind 127.0.0.1:0 and [::1]:0, relays are disabled, address lookup is cleared, port mapping is disabled, and net-report probing is reduced to its minimum.

Synchronisation is always "wait for a specific event or condition under one overall deadline" (wait_event, wait_until, 30 s). The deadline covers the probe as well as the gaps between probes: a call into a wedged agent that never answers fails the test rather than hanging the process, which is the difference between a red run and a test binary still burning a core the next day. settle() exists only for asserting that something did not happen. Ports are dynamic and directories are isolated, so tests run in parallel.

Several library instances in one process is exactly that. It is not a test of several system processes, and is not presented as one.

What is covered

#	scenario	file
1	deterministic identity: same name + secret ⇒ same space on different devices; a changed name or secret changes it; hostname, device key and restart do not	tests/identity.rs
2	four agents find each other, authenticate for real and exchange distinguishable messages; a late joiner is picked up; opaque plugin capabilities cross the control plane	tests/multi_peer.rs
3	an attacker who knows the address and the correct public NetworkId but not the secret is rejected at the handshake	tests/authentication.rs
4	one agent in two networks: statuses and messages do not mix; a session authenticated for one network cannot speak for the other; deactivating one leaves the other running	tests/network_isolation.rs
5	full stop and recreation from the same database: identity and settings survive, sessions come back automatically, a new local UDP port does not break recovery	tests/restart.rs
6	changing the secret through the library API: device identity survives, old sessions and messages get no access to the new space, and the retired space stays retired across a restart	tests/restart.rs
7	missing, corrupt and stale cache do not prevent connecting; a corrupt mandatory store is a clear error and never a fresh identity; a newer schema is refused; secrets stay out of status and Debug	tests/cache_and_state.rs
8	a dead candidate and a vanished peer do not block the others; retries are bounded and stop when the network is deactivated	tests/resilience.rs
9	wrong version, a message before authentication, a proof replayed on another connection, an oversized frame and a Hello for an inactive network are all rejected without taking the agent down	tests/authentication.rs
10	a second agent on the same state directory gets a clear error; after a clean stop the directory reopens; shutdown ends background tasks and refuses further work; independent agents coexist in one process	tests/resilience.rs
11	leaving a network frees the address for the others, says plainly when there was nobody to tell, and rejoining afterwards is not mistaken for a stale record; a wipe empties both directories and the next start is a stranger, while a directory that is not ours is refused	tests/leaving.rs, tests/cache_and_state.rs
12	a network can be joined into a running agent over the control socket and is live at once; a name with no secret resumes the one network of that name, invents one when there is none, and refuses to choose between two; two networks on one agent each get a range of their own	tests/local_control.rs, tests/network_isolation.rs, CLI unit tests

tests/wireguard.rs drives the WireGuard data plane over real iroh connections. Everything is real except the packet interface: real agents, real control plane, real data links, real WireGuard handshakes and encryption from boringtun, with an in-memory TUN device so none of it needs privileges. It covers real IPv6 packets travelling both ways through a tunnel, a three-agent mesh, a peer that sends from an address it does not own being dropped, packets for unowned addresses being counted rather than broadcast, a departing peer losing its tunnel, two networks keeping separate interfaces and keys, restart keeping the WireGuard identity, shutdown removing every interface, a forged overlay claim being rejected, and the core carrying the payload without interpreting it.

The WireGuard traffic tests constrain real QUIC to a 1200-byte path MTU with PMTU discovery disabled, so large loopback MTUs cannot hide Internet failures. Checksummed IPv4/TCP packets up to the default 1280-byte interface MTU cross in both directions without changing DF or packet contents, including through an intermediate peer. Fragment tests cover reordering, duplicates, loss, changing fragment sizes, malformed input, timeout and memory bounds. These are packet transport checks, not a claim to have run an SSH server or a host TCP stack.

Routing tests cover shortest paths, direct preference, deterministic per-flow ECMP, link loss, protocol isolation, malformed/unknown frames, bounded local queues, zero-copy transit of an owned buffer, and deliberately inconsistent tables whose loop terminates at the hop limit. A transit reader is exercised without any plugin reader, so forwarding cannot accidentally depend on one. WireGuard tests also cover flow tags queued before a handshake, queue overflow, and their preservation through encryption.

tests/wireguard.rs includes a four-agent chain A—B—C—D with only adjacent data links. Real encrypted 1280-byte TCP packets travel in both directions while the middle TUNs remain empty. A direct A—D link is enabled, then removed; the route switches back to the chain without replacing end-to-end tunnels.

Broadcast tests use limited and directed UDP game discovery packets over real WireGuard tunnels, including a missing direct link. Each willing peer gets one copy; disabling reception/origination works at runtime and unicast still works. IP-router tests cover source/destination domain checks, malformed UDP and the absence of reflection into outgoing fanout. SQLite migration tests preserve v3 identity/settings and check opt-out after reopen/rejoin. CLI tests cover default on, per-network opt-out, runtime updates and conflicting flags. Actual games and host adapter selection are not simulated by these tests.

The ignored forwarding_benchmark measures the synchronous transit routine in release mode, excluding crypto and socket I/O. Run it explicitly as described in routing.md; it has no timing threshold in the default suite.

Unit tests in crates/tsunagi-core/src/state/ cover the signed record model directly: tampering with any field breaks verification, a newer version wins while an older one never rolls back, two authors claiming one address resolve the same way no matter the merge order, one key used in two places is reported rather than silently merged, a release survives a late-arriving old claim, a bad record in a batch does not stop the rest, and allocation is deterministic, spread out, walks past everything taken and reports a full range instead of handing out a duplicate.

tests/range_conflicts.rs checks isolated first members choosing the same range, persistence across restart, preservation of legacy ranges, exclusion of equal and nested ranges, late authenticated conflicts, switching all blockers, automatic resume, cancelling a suspended network's resume and concurrent switch requests. These use real agents, loopback handshakes and SQLite; they do not change host network settings.

tests/local_control.rs covers the local control socket end to end: a client asking a running agent for status over a real Unix socket, joining and leaving a network through it, a leftover socket file being replaced while a live one is not, and the derived socket path staying short enough to bind.

crates/tsunagi/tests/network_cli.rs runs the real binary too: joining with no secret invents one, prints it in full and prints a line the other side can paste unchanged; a bare name this device already knows resumes that network instead of inventing another of the same name; and a network can be stopped and started again with its secret and its place intact. Network selection accepts exact names consistently across commands. Leave-by-name is tested with a live agent and with offline state, preserving other networks; duplicate and unknown names change nothing and full ids resolve ambiguity. Selector unit tests cover case-sensitive names, Unicode/ellipsis names, copied id prefixes and collisions between names and ids without revealing secrets.

crates/tsunagi/tests/dns_service.rs runs the real binary: the resolver comes up with no interface to attach it to, the listener is not rebuilt on the way past, a name outside every zone is refused, each network gets a zone of its own as it is joined, and the resolver can be switched on and off while the agent runs.

tests/discovery.rs also covers introductions: a device given one member's address meets every other member, and the ones it was not told about arrive as candidates that still pass the handshake like any other.

tests/discovery.rs covers the discovery contract itself: a static bootstrap candidate is enough to join, several backends compose, entries are withdrawn when a network stops, and a forgotten network stays forgotten across a restart.

Unit tests in crates/tsunagi-core/src/proto/handshake.rs cover the transcript construction itself: role separation, channel binding, identity and network binding, unambiguous encoding, and rejection under the wrong key.

Unit tests in crates/tsunagi-wg-quic/src/ cover key clamping against the RFC 7748 vector, overlay derivation, announcement validation including the address-hijack attempt, interface naming, and IP header parsing against truncated and nonsense input.

tests/end_to_end.rs is the vertical slice: persistent identity → network space → discovery → iroh → authentication → message exchange.

What the default suite does not cover is the real TUN interface, because that needs CAP_NET_ADMIN. Everything above it does run.

Mainline discovery tests use an isolated loopback mainline::Testnet. tests/mainline_socket.rs checks that idle UDP receive timeouts do not emit warnings (including Windows error 10060) and that the same node still answers a real KRPC ping afterwards. The local dependency correction is documented in vendor/mainline/PATCHES.md. tests/mainline.rs restores two agents after their DHT and disposable cache have disappeared. tests/discovery_lifecycle.rs checks that publication continues while connected, lookup resumes after isolation, and a hung backend does not block the network actor. Unit tests cover record bounds, timestamps, colliding publications, wrong secrets and the full recovery delay with a paused clock. The public Mainline/relay smoke test is ignored by default.

Not covered, and not claimed to be

Listed in sync-model.md: compaction, long partitions, NAT traversal, relay fallback and multi-host deployment. These scenarios are not established by the default loopback suite.

Debugging a test

TSUNAGI_TEST_LOG=tsunagi=debug cargo test --test multi_peer -- --nocapture

The library never installs a global subscriber; the harness opts in only when that variable is set.

The tsunagi-awg suite runs real pair negotiation and masked UDP WireGuard engines with memory TUNs. A loss wrapper drops datagrams at the carrier boundary without substituting iroh, membership, encryption or SQLite. It checks runtime loss detection, fallback and fresh profiles, in addition to restart and input validation. Tests are offline; public STUN/DPI/NAT effectiveness needs separate real-network validation.

Runtime protocol selection tests switch real AWG and iroh/QUIC carriers and end-to-end WireGuard tunnels from either endpoint-id role, carry full-size packets, keep another active network's tunnels unchanged, disable and re-enable the data plane without losing control sessions, and preserve selection after restart. CLI tests use a real agent process and control socket to exercise selection by network name, stopped networks, invalid lists, and persisted control-only configuration, offline updates, and enabling a protocol absent from the startup list. A real AWG test starts without a TUN, creates the single memory interface on enablement, and races rapid changes against pending opens. SQLite tests migrate schema 7 to 8 without changing identity, secrets or signed records, preserve order and explicit empty choices, and reject malformed stored selections. These checks use memory TUNs and do not establish GUI rendering or behavior on an actual OS interface.

tests/forget_node.rs uses real agents, handshakes and SQLite to check equal cleanup rights, network isolation, online/self/stale-version refusal, restart, recovery of a stale replica, reuse of a hostname and a returning identity's newer claim. Bounded batch delivery is checked with a two-message send queue and twelve retained markers. Retirement unit tests check signatures, proof tampering, invented future cutoffs, bounds, deterministic merge and resistance to stale snapshots. SQLite tests migrate schema 8 to 9 and inject counter-write failure to prove atomic rollback. The CLI test starts two real agent processes and exercises offline cleanup by network and hostname through the control socket. Native tray rendering and physical screen sizing are not exercised automatically.

Linux exit-client unit tests cover preservation of foreign mark bits, matching only the agent's UID and locally owned TCP/UDP sockets, exact tagged cleanup, partial IPv6 failure and mark-aware strict kernel RPF validation. The ignored overlay::hostrules::linux_exit_kernel::strict_firewall_preserves_underlay_and_application_exit_routing test creates two disposable network namespaces. It reproduces the UID-only failure with real sockets, then runs TCP/UDP IPv4/IPv6 and first inbound TCP/UDP packets, including a UDP port bound after activation, through real mark/conntrack rules with both strict kernel and Netfilter reverse-path checks. It verifies application exit routing, IPv6 blocking, repeated apply and cleanup with foreign rules at Tailscale's priority retained. It requires root, iproute2, iptables/ip6tables and Python; it does not run in the default suite or claim to test the entire mesh inside these namespaces.

Devices-window tests render the egui grid at multiple widths with its real scroll area. They check stretching and shrinking without an expanding first column, separate handshake/control RTT values, unknown measurements and short end-to-end hints that do not substitute the carrier protocol. This does not establish native desktop rendering on every platform.

VIP tests cover signed multi-writer backend merging, stale draft rejection, permanent deletion and replay, separate active/deleted capacity, and mandatory SQLite migration/counter rollback. Real agent instances verify equal editing rights, network isolation, restart and recovery without the creator. Packet checks exercise RTT selection, port translation, UDP failover, a local backend, fragment reassembly/checksums and ICMP quoted-header/PMTU translation. TCP check recovery uses real loopback listening sockets. The WireGuard suite moves VIP TCP/UDP packets through real encrypted tunnels, rejects forged VIP sources and checks UDP failover after the selected agent stops. CLI/DNS tests run the real binary, configure independent listeners, disable checks, reject an invalid change atomically and observe DNS creation/deletion. GUI tests cover draft isolation and form rendering. These do not establish live host TUN behavior or native desktop behavior on Linux, macOS or Windows. See VIP semantics.

Tray layout tests render bounded table rows with long names, online/offline members, Forget buttons, large counters and enlarged fonts. They assert text stays within its own row and that narrow windows scroll without expanding the viewport. VIP overview tests render live tables and egui scenes at multiple widths, click a diagram backend, and distinguish the new-flow selection from traffic on an existing binding. Sampling checks cover idle open flows, stale observations, counter resets, recreated listeners and removal of cached histories. These are headless egui checks, not native desktop screenshots.