fix: one node per gateway, container bind docs, and two test races

README.md

@@ -99,6 +99,13 @@ sudo apt install ros-jazzy-ros2-medkit-gateway   # or ros-humble- / ros-lyrical-
 ros2 launch ros2_medkit_gateway bringup.launch.py
 ```
 
+> [!IMPORTANT]
+> `bringup.launch.py` binds the REST API to `127.0.0.1`. Inside a container without
+> `--network host` that is the container's own loopback, and the host gets
+> `Connection reset by peer` or an empty reply on the published port. Pass `server_host:=0.0.0.0`
+> and publish the port. Publishing it as `-p 127.0.0.1:8080:8080` keeps the API, which has no
+> authentication by default, off the LAN.
+
 > [!TIP]
 > That is the whole setup. It auto-discovers every node, topic, service and action and starts
 > emitting structured faults over REST - no instrumentation, no changes to your stack. Prefer
@@ -157,8 +164,11 @@ docker run -p 3000:80 ghcr.io/selfpatch/ros2_medkit_web_ui:latest
 ```
 
 The browser calls the gateway from a different origin, so the gateway must allow that origin via
-CORS (the prebuilt gateway Docker image enables it; for a native bringup set
-`cors.allowed_origins`). See the [web UI tutorial](https://selfpatch.github.io/ros2_medkit/tutorials/web-ui.html).
+CORS. The Docker image and `bringup.launch.py` allow `http://localhost:3000` and
+`http://localhost:5173`. The origin must match exactly: `http://127.0.0.1:3000` is not
+`http://localhost:3000`. To allow other origins, pass the full list, for example
+`cors_allowed_origins:=http://localhost:3000,http://127.0.0.1:3000` (it replaces the defaults).
+See the [web UI tutorial](https://selfpatch.github.io/ros2_medkit/tutorials/web-ui.html).
 
 </details>
 

docs/api/rest.rst

@@ -229,9 +229,9 @@ Server Capabilities
               "/powertrain/engine/rpm_sensor",
               "/ros2_medkit_gateway",
               "/_param_client_node",
-              "/ros2_medkit_gateway_fault_clients",
-              "/ros2_medkit_gateway_lifecycle_state_reader",
-              "/ros2_medkit_gateway_sub"
+              "/_ros2_medkit_gateway_fault_clients",
+              "/_ros2_medkit_gateway_lifecycle_state_reader",
+              "/_ros2_medkit_gateway_sub"
             ],
             "peer_names": []
           }

docs/config/discovery-options.rst

@@ -81,7 +81,9 @@ Internal Node Filtering
 When ``filter_internal_nodes`` is true (the default), ROS 2 nodes whose names
 start with an underscore (``_``) are excluded from the entity tree. This
 filters out ROS 2 internal infrastructure nodes such as ``_ros2cli_*``,
-``_param_client_node``, and similar system nodes that should not appear as
+``_param_client_node``, the gateway's own helper nodes
+(``_<gateway name>_fault_clients``, ``_<gateway name>_lifecycle_state_reader``,
+``_<gateway name>_sub``), and similar system nodes that should not appear as
 SOVD entities. The filter applies to both locally discovered Apps and
 peer-discovered Apps (after stripping the peer prefix).
 

docs/getting_started.rst

@@ -46,8 +46,8 @@ You should see:
 
 .. code-block:: text
 
-   [gateway_node]: REST server starting on http://127.0.0.1:8080
-   [gateway_node]: REST server started successfully
+   [ros2_medkit_gateway]: Configuration: REST API at 127.0.0.1:8080, backstop refresh interval: 30000ms
+   [ros2_medkit_gateway]: ROS 2 Medkit Gateway ready on HTTP://127.0.0.1:8080
 
 **Terminal 2 - Start demo nodes:**
 

docs/troubleshooting.rst

@@ -73,6 +73,34 @@ For network access, set host to 0.0.0.0:
 
    ros2 launch ros2_medkit_gateway gateway.launch.py server_host:=0.0.0.0
 
+**"Connection reset by peer" or "Empty reply from server" from a container**
+
+.. code-block:: text
+
+   curl: (56) Recv failure: Connection reset by peer
+   curl: (52) Empty reply from server
+
+The gateway runs in a container on a bridge network with a published port
+(``-p 8080:8080``), ``curl`` inside the container works, and ``curl`` on the host
+fails. The gateway listens on the container's loopback: ``bringup.launch.py`` and
+``gateway.launch.py`` default to ``server_host:=127.0.0.1``. The startup log shows
+the bind address:
+
+.. code-block:: text
+
+   Configuration: REST API at 127.0.0.1:8080, backstop refresh interval: 30000ms
+
+Bind all interfaces inside the container:
+
+.. code-block:: bash
+
+   ros2 launch ros2_medkit_gateway bringup.launch.py server_host:=0.0.0.0
+
+Publish the port as ``-p 127.0.0.1:8080:8080`` to keep the API off the LAN;
+authentication is off by default. With ``--network host`` the default bind is
+enough, because the container shares the host's loopback. The Docker image's
+default command already binds ``0.0.0.0``.
+
 **Topic data returns empty or timeout**
 
 Possible causes:
@@ -163,9 +191,19 @@ For development, try ``network_mode: host``.
 
 **Web UI can't connect to gateway in container**
 
-1. Gateway must listen on ``0.0.0.0``, not ``127.0.0.1``
-2. CORS must allow the UI origin
-3. Port must be exposed in docker-compose
+1. Gateway must listen on ``0.0.0.0``, not ``127.0.0.1``. The image's default
+   command does; a launch file needs the argument:
+
+   .. code-block:: bash
+
+      ros2 launch ros2_medkit_gateway bringup.launch.py server_host:=0.0.0.0
+
+2. CORS must allow the UI origin exactly as the browser sends it (scheme, host
+   and port): ``http://127.0.0.1:3000`` is not ``http://localhost:3000``. The
+   defaults are ``http://localhost:3000`` and ``http://localhost:5173``. Pass
+   ``cors_allowed_origins:=<origin>,<origin>`` to the launch file for others; it
+   replaces the defaults, so list every origin you need.
+3. Port must be published (``-p 8080:8080`` or ``ports`` in docker-compose)
 
 Fault Manager Issues
 --------------------

docs/tutorials/bringup.rst

@@ -48,6 +48,11 @@ launch arguments:
    * - ``server_port``
      - ``8080``
      - Gateway REST API port.
+   * - ``cors_allowed_origins``
+     - ``http://localhost:3000,http://localhost:5173``
+     - Comma-separated browser origins allowed to call the gateway. Each must
+       match exactly (``http://127.0.0.1:3000`` is not ``http://localhost:3000``).
+       A value replaces the defaults; empty disables CORS.
    * - ``enable_fault_manager``
      - ``true``
      - Start the fault_manager node.
@@ -62,6 +67,14 @@ launch arguments:
      - Start the diagnostic_bridge (``/diagnostics`` -> faults). Opt-in, for
        legacy ``diagnostic_updater`` publishers.
 
+.. note::
+
+   In a container without ``--network host``, ``127.0.0.1`` is the container's
+   own loopback: the published port answers with ``Connection reset by peer``
+   or an empty reply. Run ``bringup.launch.py server_host:=0.0.0.0`` and publish
+   the port as ``-p 127.0.0.1:8080:8080`` to keep the API (no authentication by
+   default) off the LAN.
+
 Verify
 ------
 

docs/tutorials/docker.rst

@@ -67,10 +67,28 @@ Test the gateway:
 Custom Configuration
 --------------------
 
-The default configuration listens on ``0.0.0.0:8080``. CORS is enabled for the
-default web UI origins (``http://localhost:3000`` and ``http://localhost:5173``)
-so the web UI works out of the box; add your own UI origin(s) as needed (see
-`CORS for Web UI`_ below). To use a custom configuration, mount a params file:
+The image's default command runs the gateway alone with
+``/etc/ros2_medkit/params.yaml``, which listens on ``0.0.0.0:8080``. CORS is
+enabled for the default web UI origins (``http://localhost:3000`` and
+``http://localhost:5173``) so the web UI works out of the box; add your own UI
+origin(s) as needed (see `CORS for Web UI`_ below).
+
+.. note::
+
+   A launch file run in the container binds ``127.0.0.1`` unless told
+   otherwise. Without ``--network host`` that is the container's own loopback,
+   and the published port answers with ``Connection reset by peer`` or an empty
+   reply. Pass ``server_host:=0.0.0.0``:
+
+   .. code-block:: bash
+
+      docker run -p 127.0.0.1:8080:8080 ghcr.io/selfpatch/ros2_medkit-jazzy:latest \
+        ros2 launch ros2_medkit_gateway bringup.launch.py server_host:=0.0.0.0
+
+   ``127.0.0.1:8080:8080`` publishes the port on the host's loopback only, which
+   keeps the API (no authentication by default) off the LAN.
+
+To use a custom configuration, mount a params file:
 
 .. code-block:: bash
 
@@ -227,6 +245,11 @@ writes. Add your own UI origin(s):
            - "http://localhost:3000"
            - "https://my-dashboard.example.com"
 
+The origin must match what the browser sends exactly: ``http://127.0.0.1:3000``
+is not ``http://localhost:3000``. When the container runs a launch file, pass
+the origins as ``cors_allowed_origins:=http://localhost:3000,http://127.0.0.1:3000``;
+the argument replaces the defaults.
+
 Health Checks
 -------------
 

docs/tutorials/web-ui.rst

@@ -105,8 +105,11 @@ Connecting to ros2_medkit
 
 .. tip::
 
-   If the gateway runs on a different host, ensure CORS is configured.
-   See :doc:`/config/server` for CORS settings.
+   The web UI is a different origin than the gateway, so the gateway must allow
+   it via CORS. The launch files and the Docker image allow
+   ``http://localhost:3000`` and ``http://localhost:5173``, matched exactly:
+   ``http://127.0.0.1:3000`` is rejected. See :doc:`/config/server` for CORS
+   settings, or pass ``cors_allowed_origins:=`` to the launch file.
 
 Using the Interface
 -------------------
@@ -182,21 +185,24 @@ Run both gateway and web UI together:
 .. code-block:: yaml
 
    # docker-compose.yml
-   version: '3.8'
    services:
      gateway:
-       image: ros:jazzy
-       command: >
-         bash -c "source /opt/ros/jazzy/setup.bash &&
-                  ros2 launch ros2_medkit_gateway gateway.launch.py server_host:=0.0.0.0"
-       ports:
-         - "8080:8080"
+       image: ghcr.io/selfpatch/ros2_medkit-jazzy:latest
+       command: ros2 launch ros2_medkit_gateway bringup.launch.py
        network_mode: host
+       ipc: host
 
      web_ui:
        image: ghcr.io/selfpatch/ros2_medkit_web_ui:latest
        ports:
-         - "80:80"
+         - "3000:80"
+
+The gateway shares the host network, so it joins the robot's DDS graph and its
+default bind ``127.0.0.1:8080`` is the host's loopback. Open
+``http://localhost:3000``: that origin is allowed by default, while
+``http://127.0.0.1:3000`` is not. On a bridge network instead, run
+``bringup.launch.py server_host:=0.0.0.0`` and publish ``8080`` (see
+:doc:`/troubleshooting`).
 
 Docker Image Tags
 -----------------

src/ros2_medkit_fault_manager/test/test_fault_manager.cpp

@@ -1060,7 +1060,7 @@ class FaultEventPublishingTest : public ::testing::Test {
     std::string events_topic = ns + "/fault_manager/events";
     auto qos = rclcpp::QoS(100).reliable().durability_volatile();
     event_subscription_ =
-        test_node_->create_subscription<FaultEvent>(events_topic, qos, [this](const FaultEvent::SharedPtr msg) {
+        test_node_->create_subscription<FaultEvent>(events_topic, qos, [this](const FaultEvent::ConstSharedPtr & msg) {
           received_events_.push_back(*msg);
         });
 
@@ -1206,7 +1206,7 @@ TEST_F(FaultEventPublishingTest, NewFaultPublishesConfirmedEvent) {
 
   // Wait for event to arrive (polling, robust under CPU contention)
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Verify EVENT_CONFIRMED was published
@@ -1221,7 +1221,7 @@ TEST_F(FaultEventPublishingTest, UpdateExistingFaultPublishesUpdatedEvent) {
   // Report a new fault first
   ASSERT_TRUE(call_report_fault("TEST_FAULT_2", Fault::SEVERITY_WARN, "/test_node1"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Clear received events
@@ -1230,7 +1230,7 @@ TEST_F(FaultEventPublishingTest, UpdateExistingFaultPublishesUpdatedEvent) {
   // Report same fault again - should trigger EVENT_UPDATED
   ASSERT_TRUE(call_report_fault("TEST_FAULT_2", Fault::SEVERITY_ERROR, "/test_node2"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Verify EVENT_UPDATED was published (severity/sources changed; still one occurrence)
@@ -1244,7 +1244,7 @@ TEST_F(FaultEventPublishingTest, ClearFaultPublishesClearedEvent) {
   // Report a fault first
   ASSERT_TRUE(call_report_fault("TEST_FAULT_3", Fault::SEVERITY_ERROR, "/test_node"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Clear received events
@@ -1253,7 +1253,7 @@ TEST_F(FaultEventPublishingTest, ClearFaultPublishesClearedEvent) {
   // Clear the fault
   ASSERT_TRUE(call_clear_fault("TEST_FAULT_3"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Verify EVENT_CLEARED was published
@@ -1280,7 +1280,7 @@ class HealingFaultEventPublishingTest : public FaultEventPublishingTest {
 TEST_F(HealingFaultEventPublishingTest, HealPublishesClearedEventSoStreamConsumersSeeTheEnd) {
   ASSERT_TRUE(call_report_fault("HEAL_ME", Fault::SEVERITY_ERROR, "/test_node"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
   received_events_.clear();
 
@@ -1323,7 +1323,7 @@ TEST_F(FaultEventPublishingTest, EventContainsCorrectTimestamp) {
 
   ASSERT_TRUE(call_report_fault("TEST_FAULT_4", Fault::SEVERITY_WARN, "/test_node"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   auto after = fault_manager_->now();
@@ -1339,7 +1339,7 @@ TEST_F(FaultEventPublishingTest, EventContainsCorrectTimestamp) {
 TEST_F(FaultEventPublishingTest, EventContainsFullFaultData) {
   ASSERT_TRUE(call_report_fault("FULL_DATA_TEST", Fault::SEVERITY_CRITICAL, "/sensor/temperature"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   ASSERT_EQ(received_events_.size(), 1u);
@@ -1362,7 +1362,7 @@ TEST_F(FaultEventPublishingTest, TimestampUsesWallClockNotSimTime) {
 
   ASSERT_TRUE(call_report_fault("WALL_CLOCK_TEST", Fault::SEVERITY_WARN, "/test_node"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   auto wall_after = std::chrono::system_clock::now();
@@ -1390,7 +1390,7 @@ TEST_F(FaultEventPublishingTest, GetFaultReturnsExpectedFault) {
   // Report a fault first
   ASSERT_TRUE(call_report_fault("GET_FAULT_TEST", Fault::SEVERITY_ERROR, "/test_node"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Get fault via service
@@ -1414,7 +1414,7 @@ TEST_F(FaultEventPublishingTest, GetFaultReturnsEnvironmentData) {
   // Report a fault
   ASSERT_TRUE(call_report_fault("ENV_DATA_TEST", Fault::SEVERITY_WARN, "/sensor/temp"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   auto response = call_get_fault("ENV_DATA_TEST");
@@ -1434,7 +1434,7 @@ TEST_F(FaultEventPublishingTest, GetFaultReturnsExtendedDataRecords) {
   // Report fault twice to have first and last occurrence timestamps differ
   ASSERT_TRUE(call_report_fault("EDR_TEST", Fault::SEVERITY_ERROR, "/node1"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
   ASSERT_TRUE(call_report_fault("EDR_TEST", Fault::SEVERITY_ERROR, "/node2"));
   ASSERT_TRUE(spin_until([this]() {
@@ -1482,7 +1482,7 @@ TEST_F(FaultEventPublishingTest, ListFaultsForEntityEmptyResult) {
   // Report faults from a different entity
   ASSERT_TRUE(call_report_fault("SOME_FAULT", Fault::SEVERITY_ERROR, "/some/other_entity"));
   ASSERT_TRUE(spin_until([this]() {
-    return received_events_.size() >= 1;
+    return !received_events_.empty();
   }));
 
   // Query faults for non-existent entity
@@ -1883,6 +1883,10 @@ class SnapshotCooldownTest : public ::testing::Test {
     spin_thread_ = std::thread([this]() {
       executor_.spin();
     });
+    // A cancel() that lands before spin() starts is lost, and join() would block.
+    while (!executor_.is_spinning()) {
+      std::this_thread::yield();
+    }
 
     ASSERT_TRUE(report_client_->wait_for_service(std::chrono::seconds(5)));
     ASSERT_TRUE(clear_client_->wait_for_service(std::chrono::seconds(5)));

src/ros2_medkit_gateway/CMakeLists.txt

@@ -210,6 +210,7 @@ add_library(gateway_ros2 STATIC
   src/plugins/plugin_loader.cpp
   src/plugins/plugin_manager.cpp
   src/ros2_common/callback_groups.cpp
+  src/ros2_common/helper_node.cpp
   src/ros2_common/ros2_subscription_executor.cpp
   src/ros2_common/ros2_subscription_slot.cpp
   src/script_manager.cpp
@@ -882,6 +883,11 @@ if(BUILD_TESTING)
   target_link_libraries(test_ros2_lifecycle_state_reader gateway_ros2)
   medkit_target_dependencies(test_ros2_lifecycle_state_reader rclcpp lifecycle_msgs)
 
+  # Helper nodes keep their own name under a process-wide __node remap
+  medkit_add_gtest(test_helper_node test/test_helper_node.cpp)
+  target_link_libraries(test_helper_node gateway_ros2)
+  medkit_target_dependencies(test_helper_node rclcpp)
+
   # Private client nodes torn down after a first graph wait that follows rclcpp::shutdown()
   medkit_add_gtest(test_graph_listener_join test/test_graph_listener_join.cpp)
   target_link_libraries(test_graph_listener_join gateway_ros2)