logos-messaging-module/tests/e2e

logos-delivery-module e2e tests

End-to-end checks that run the built delivery_module + liblogosdelivery inside real logoscore daemon containers and drive them over RPC. Per PR: red on regressions in delivery_module / liblogosdelivery / the codegen+RPC path.

Mock-based C++ unit tests (and the in-process C++ integration tests) live in ../tests/. This is a separate suite that runs the real stack across a container boundary, which the C++ tests can't.

Architecture

Two delivery_module daemons in a shared docker network, peered directly via staticnodes — no bootstrap node. Node A starts first; node B is configured with A's container-routable multiaddr and dials it; gossipsub/relay handles the rest.

docker network "logosdelivery-e2e-<uuid>"  (172.31.0.0/16)
  ┌────────────────────────────┐        ┌────────────────────────────┐
  │ logoscore-a  (node A)       │◀──────▶│ logoscore-b  (node B)       │
  │ relay, cluster 198, shard 0 │ relay  │ staticnodes=[A multiaddr]   │
  └────────────────────────────┘        └────────────────────────────┘

This mirrors the proven two-node relay setup in logos-delivery-interop-tests (tests/wrappers_tests/test_send_relay_propagation.py, scenario S06): relay-only, single shard, discv5 off. The single-node lifecycle tests use one daemon each.

What's tested

Test	Daemons	Asserts
test_create_start_and_stop_node	1	load_module → createNode → start → stop all succeed over RPC
test_createNode_twice_rejected	1	second createNode returns success=false
test_query_and_subscribe_methods	1	queries (getAvailableConfigs/getAvailableNodeInfoIDs/getNodeInfo) + subscribe/unsubscribe round-trip
test_two_nodes_propagation	2	gate — sender sees messagePropagated for its requestId
test_bidirectional_propagation	2	A→B and B→A both propagate
test_two_nodes_message_received	2	receiver sees messageReceived (see below)

We assert messagePropagated, not messageSent: with relay-only / no store peer, messageSent never fires (interop S06 vs S07).

Event payload shape (read before extending receive assertions)

delivery_module's public methods are synchronous — client.call(...) returns the StdLogosResult {success, value, error} directly; there are no per-call result events. Only message delivery is async, via the typed events messageReceived / messagePropagated / messageSent / messageError / connectionStateChanged.

The logoscore CLI serializes a typed event as {"event": <name>, "data": {"arg0": <1st arg>, "arg1": <2nd arg>, ...}} — each codegen event arg lands at argN by position (so messagePropagated.requestId = arg0, messageReceived.contentTopic = arg1). parse_event / event_arg in libs/helpers.py read that.

messageReceived fires and its contentTopic (arg1) is asserted. The vector<uint8_t> payload arg's serialization is not yet asserted — if you extend the receive checks to the payload, read messageReceived payload: … from the logs and confirm the decoding in libs/helpers.py first.

Prerequisites

• docker available on host.
• LOGOSCORE_IMAGE env: logoscore:smoke-portable, built once from logos-co/logos-logoscore-py/tests/docker_smoke/build_smoke_image.sh.
• LOGOS_MODULES_DIR env: a directory containing delivery_module/manifest.json (the output of nix build .#install-portable → result/modules).
• A host-side logoscore CLI on PATH (the client shells out to it). Build it from logos-co/logos-logoscore-cli pinned to the same commit as the smoke image.

If any of these is missing, all e2e tests skip with a clear reason.

Running locally — Linux

# 1. Build the module in install-portable layout.
nix build .#install-portable
export LOGOS_MODULES_DIR=$PWD/result/modules

# 2. Build the logoscore smoke image (one-time) + CLI, pinned together.
git clone https://github.com/logos-co/logos-logoscore-py.git /tmp/logoscore-py
cd /tmp/logoscore-py && git checkout aa45db52
bash tests/docker_smoke/build_smoke_image.sh
export LOGOSCORE_IMAGE=logoscore:smoke-portable
cd -
nix build --out-link /tmp/logoscore-cli \
  "github:logos-co/logos-logoscore-cli/5a1cf746e841e8e8a8fd48333cff55bd12c45c52#cli"
export PATH="/tmp/logoscore-cli/bin:$PATH"

# 3. Install deps + run.
cd tests/e2e
pip install -r requirements.txt
pytest -v

Running locally — macOS

Docker-mode on macOS (arm64) needs a Linux remote builder for nix build .#install-portable (the Linux delivery_module_plugin.so + bundled dylibs), which isn't shipped here. Use CI to validate, or a Linux host.

Configuration schema

Node configs are built by libs/helpers.make_delivery_config() and follow the createNode WakuNodeConf JSON schema (see ../README.md#node-configuration-createnode):

Field	Value here	Why
relay	true	gossipsub relay path under test
clusterId	198	isolated test cluster (interop's DEFAULT_CLUSTER_ID)
numShardsInNetwork	1	single shard → every content topic maps to shard 0
discv5Discovery	false	peering is explicit via staticnodes, not discovery
staticnodes	[A multiaddr] (node B only)	direct dial of node A
store / filter / lightpush	false	not exercised here

The cluster + shard must match across both nodes (they do — same config builder). A peer is reached via getNodeInfo("MyMultiaddresses") with the container IP substituted in (helpers.node_multiaddr), since the advertised address is 0.0.0.0.

Troubleshooting

Daemon container logs are saved to $E2E_LOG_DIR (default /tmp) on teardown, and uploaded as the e2e-test-logs artifact in CI on failure. Set LOGOSCORE_PY_FORWARD_OUTPUT=1 to also mirror CLI + daemon output into pytest logs.

CI

.github/workflows/ci.yml has an e2e-tests job (needs: build-and-test, ubuntu-latest) that builds .#install-portable, builds the smoke image + CLI, runs pytest -v, and uploads logs on failure. build-and-test pre-warms the install-portable cachix cache so the e2e job hits it warm.