Skip to content

feat(jmex): add AgileMaster full-body plugin and G1 teleop example - #1250

Open
jmex-sw wants to merge 5 commits into
NVIDIA:mainfrom
jmex-sw:jmex-plugin-full-body
Open

jmex-sw wants to merge 5 commits into
NVIDIA:mainfrom
jmex-sw:jmex-plugin-full-body

Conversation

@jmex-sw

@jmex-sw jmex-sw commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

Description

Adds support for the j-mex AgileMaster motion capture suit: a device plugin that publishes the
operator's body pose as FullBodyPose under a new body.jmex vendor, the core registration that
lets Isaac Capture read it, and an Isaac Lab example that drives a Unitree G1 from it. None of this
is upstream yet; the plugin, the example and their tests are all new.

AgileMaster has three parts: the Mocap Suit the operator wears; MOXI Player, which solves the suit
onto a skeleton and streams it on a numbered channel; and the MOXI Receiver SDK, which the plugin
links against.

The plugin (src/plugins/jmex/) receives one Player channel and fills every BodyJoint slot
from the standard human skeleton loaded in Player, one slot per joint, in OpenXR axes. It pushes one
FullBodyPose per frame, on the jmex_full_body collection by default. A slot with no joint, or
with no usable value in a frame, goes out with is_valid false; nothing is inferred from
neighboring joints. The raw device clock carries the frame's time from Player
(--device-clock=local uses the pickup time instead). If the CloudXR runtime restarts underneath
it, the plugin re-establishes its session and carries on. It is configured by flags only. An
argument it cannot use, such as a misspelt flag, an empty value or a channel outside 0–255, stops
it with exit code 1 and names the argument; --help lists the flags and their defaults. Launched by
TeleopSession, it listens and advertises on loopback only, so only a MOXI Player on the same
machine can pair; its README says what a Player on another machine needs. It builds only with
BUILD_PLUGIN_JMEX on (default OFF), runs on Linux x86_64 and needs the MOXI Receiver SDK, which
find_package locates through CMAKE_PREFIX_PATH, MOXIReceiverSDK_DIR or the plugin directory.
A MOXIReceiverSDK_DIR that is set is the only place searched, and configure stops if it holds no
SDK. With the option on and no SDK found, or on a target other than x86_64, configure warns and
skips the plugin executable. The plugin's SDK-free core library, which the tests link, builds on
Linux with or without the option; the rest of the tree builds as before.

Where the conversion happens. Player streams its own skeleton. The plugin maps it onto the
FullBodyPose slots and into OpenXR axes (body_joint_mapping, openxr_convention); Player does
not emit FullBodyPose. There are two reasons for this. The change of axes was measured in this
tree and is pinned by the plugin's tests, next to the code that consumes the slots. Done in Player,
those tests would live where this repository's CI never runs them. And Player does not have to
produce Isaac Capture's internal schema, so a change to FullBodyPose stays in this tree instead of
waiting on a Player release.

Registration in core. Outside the new directories, every change is an addition; no existing line
changes.

Where What
src/core/live_trackers/cpp/live_full_body_tracker_jmex_impl.{hpp,cpp} New: the live tracker for body.jmex. It reads the plugin's collection and takes one vendor parameter, collection_id
src/core/live_trackers/cpp/CMakeLists.txt Those two files in the library's source list
live_deviceio_factory.{hpp,cpp} A factory method and its try_create_* function; one row in k_tracker_dispatch, after the existing full-body vendors, so FullBodyTracker's default vendor is unchanged; one branch in validate_vendor_selections
tests/cpp/core/live_trackers/test_vendor_validation.cpp Four SECTIONs for body.jmex, beside the existing vendors'
CMakeLists.txt BUILD_PLUGIN_JMEX, default OFF, after the other device plugins' options
src/plugins/CMakeLists.txt add_subdirectory(jmex) with the plugins that are always entered; off Linux the plugin skips itself, and on Linux it builds its core library, and the plugin executable too when BUILD_PLUGIN_JMEX is on and the SDK is found on x86_64
tests/cpp/plugins/CMakeLists.txt add_subdirectory(jmex) for the plugin's tests, when its core library is built
tests/python/examples/CMakeLists.txt add_subdirectory(jmex) for the example's tests
cmake/ClangFormat.cmake An SDK unpacked into the plugin directory, under any versioned name, is excluded from the format check
docs/source/_data/devices.yaml A row for the Ecosystem page

The example (examples/jmex/, package isaaccapture_examples.jmex) has two parts, and a
plugin module that holds the plugin's name, ids and defaults and launches it from
install/plugins:

  • jmex_g1_tasks registers Isaac-PickPlace-Locomanipulation-G1-Jmex-Abs-v0, which inherits Isaac
    Lab's LocomanipulationG1EnvCfg (Isaac Lab 3.0.0, 3.0.0 EA or 3.0.0 beta2) and takes its action from the
    operator; nothing in Isaac Lab is modified. Its pipeline:

    FullBodySource (vendor body.jmex)
      -> JmexG1Retargeter                   frame task poses
    LocomotionFixedRootCmdRetargeter        standing command
      -> TensorReorderer                    poses, hand_joint_dim fingers left at zero, command
        -> OutputCombiner                   the task's action
    

    This is upper-body teleop. Each arm is rebuilt segment by segment onto the G1's own shoulder, so
    the operator's arm angles carry over at the robot's arm lengths, and the elbows and shoulders get
    IK targets from that same chain. The wrist's orientation is the operator's hand, carried through
    a forearm frame. The operator's torso turns the robot's shoulders about its waist, so leaning,
    bending sideways and twisting come through; their heading does not, so the operator can turn to
    face a screen without moving the robot. Only a frame that carries every slot drives the robot.
    The state kept across frames is the retargeter's; the source node is upstream's.

  • record_jmex_full_body records the same stream to a standard full-body MCAP
    (core.FullBodyPoseRecord on the full_body channel), which the shared replay_full_body script
    plays back unmodified.

Type of change

  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to change)
  • Documentation update

Testing

The tests in this PR need neither the SDK nor BUILD_PLUGIN_JMEX. A configure on Linux,
CI's included, registers them without either, and ctest -L jmex selects 25: the plugin's
22 Catch2 cases (argument parsing and --help, slot mapping, change of axes, device clock,
one frame becoming one FullBodyPose, session recovery after a failed push), which link the
plugin's SDK-free core library; and the example's three Python tests, whose 98 pytest cases
cover the retargeter, including the geometry of the rebuilt arm and the orientation carried through
the forearm frame, the composed action's length and order, the plugin launcher, and the recording
script's --help without Isaac Lab. The Python tests also need BUILD_EXAMPLES and BUILD_VIZ. The
body.jmex sections of the vendor validation test are in core and run under
ctest -R 'vendor validation' with or without the option.

Local gate, Ubuntu x86_64. We ran the same Python and build-type matrix as the x64 entries of
build-ubuntu (Python 3.11, 3.12 and 3.13, each in Debug and Release), with BUILD_PLUGIN_JMEX off
as in CI and without the SDK, each from an empty tree, twice: once with CI's toolchain (GCC 11.4, CMake
3.31.6, uv 0.12.13, clang-format 14.0.0) and once with GCC 13.3. Local builds left
BUILD_OVXR_CAMERA_QUADS and BUILD_PLUGIN_OAK_CAMERA disabled; CI enables both. In all twelve:

  • configure printed nothing of the plugin, and built its core library and tests but not the executable;
  • the build compiled 828 objects, and its clang-format check passed over 437 files;
  • the 25 jmex tests and the 3 vendor validation cases passed;
  • the 10 of this change's 12 .cpp files that build without the SDK compiled without warnings (the
    other two are covered below);
  • of the 720 tests in the full ctest, 710 passed, 9 skipped and
    1 failed: an upstream GPU test (VkContext with an explicit device index) that fails on this
    workstation's GPU setup and is unrelated to this change; CI's runners have no GPU and skip it.

With the option on and no SDK, configure warned and skipped the executable, and registered the same
25 tests; with SDK 0.8.1 it also built and installed the executable. With MOXIReceiverSDK_DIR
pointing at an empty directory, configure stopped with the error, also with a usable SDK on
CMAKE_PREFIX_PATH or in the plugin directory; it stopped likewise at SDK 0.8.0, and at an SDK moved
away after a configure had found it. Each stopped again on a second configure.

The two plugin sources that compile only with the SDK were built against SDK 0.8.1 with GCC 13 and
syntax-checked with GCC 11, without warnings. With CI's toolchain we also repeated the Release
install with auditwheel repair, the pip-editable-install job (its import check, and its
wheel-content tests: 10 passed) and the configure of test-viz-sanitizers.
SKIP=check-copyright-year pre-commit run --all-files, make current-docs (-W),
scripts/check_doc_refs.py --all, ruff, and clang-format 14 on the changed files all pass. Windows
and arm64 were not built locally.

Since this PR comes from a fork, its workflows wait for a maintainer's approval. The same commit has
already been through the Build Ubuntu workflow on our fork, where all twelve build-ubuntu entries
(x64 and arm64) passed, along with DCO and pre-commit:
https://github.com/jmex-sw/IsaacTeleop/actions/runs/44007964864. Those entries build the core
library and run the 25 jmex tests, on x64 and arm64 alike.

With the device. With the current pipeline and isaaccapture 1.8 in Isaac Lab's Python, a
recorded Player stream replayed over loopback drove the task headless on all three Isaac Lab
versions: each of 900 steps was computed from a full body pose, at about 12 steps/s on beta2, 11
on EA and 10 on 3.0.0, and the action-length check passed. On Isaac Lab 3.0.0 with Kit's window
open and MOXI Player for Robot on the same machine, the plugin the task launched paired over
loopback and the G1 followed. When Player disconnected, the G1 held its last pose, and it followed
again once Player paired again. Ctrl-C ended the plugin on its counters line, with no frames
dropped.

On 2026-11-06, with SDK 0.8.1 and MOXI Player for Robot on the same machine, the plugin with
plugin.yaml's loopback flags paired and published within 6 s in both of two runs, its TCP server
listening only on 127.0.0.1. With Player on a second machine on the same subnet, the README's
settings for that case paired within 5 s in both of two runs, and plugin.yaml's loopback flags
kept that Player out: it did not connect in 40 s.

The runs below were made on earlier heads of this PR. Since then the skeleton mapping is unchanged,
and the example's retargeting moved onto FullBodySource and the upstream nodes with the same
geometry; its geometry tests changed only in how they feed the input. On 2026-10-30, driven live by
an operator wearing the upper-body part of the suit, with MOXI Player on the same machine (Isaac
Lab 3.0.0 beta2 on Isaac Sim 6.0.1, XR running, retargeting in sync mode), the G1's waist followed
lean, side bend and twist with a correlation of 0.97–0.99, 90–270 ms behind. The operator went past
the waist's limits in lean and twist, where the waist stops following. An earlier run in the full
suit, on 2026-09-23 with pipelined retargeting, gave 0.99 at the same latencies, and turning on the
spot through ±90° moved the G1's yaw less than 12°. Replayed headless from a recording (rerun on
2026-10-30), the waist followed all three with a correlation of 1.00, one or two physics steps
behind. On 2026-10-30, on Isaac Lab 3.0.0 EA (Isaac Sim 6.1.0), the G1 followed a replayed
recording with Kit's window open and the reference overlay drawn.

Running it yourself. Isaac Lab needs this repository's isaaccapture installed in its Python:
no released package carries the body.jmex vendor yet, and Isaac Lab pins the package under its
previous name. The example README has the steps. The SDK and MOXI Player are not distributed with
this PR.

Checklist

  • I have read and understood the contribution guidelines
  • I have run the linter and formatter with SKIP=check-copyright-year pre-commit run --all-files
  • I have made corresponding changes to the documentation
  • I have added tests that prove my fix/feature works (or explained why not)
  • I have signed off all my commits (git commit -s) per the DCO

@github-actions

Copy link
Copy Markdown
Contributor

📝 Docs preview is not auto-deployed for fork PRs.

A maintainer with write access to NVIDIA/IsaacCapture can deploy a preview by
commenting /preview-docs on this PR. Once deployed, the preview
will live at:

https://nvidia.github.io/IsaacCapture/preview/pr-1250/

@coderabbitai

coderabbitai Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

Review in Change Stack →

No actionable comments were generated in the recent review.

Walkthrough

This change adds a j-mex plugin that receives AgileMaster full-body poses and publishes them through OpenXR. It adds body.jmex live-tracker support and a G1 teleoperation example that converts received poses into robot action targets. It also adds an MCAP recording example, build and test configuration, and documentation for the plugin and examples.

Estimated code review effort: 4 (Complex) | ~60 minutes

Sequence Diagram(s)

sequenceDiagram
  participant PluginMain
  participant ReceiverSession
  participant FullBodyPublisher
  participant OpenXRTransport
  PluginMain->>ReceiverSession: Poll for a new frame
  ReceiverSession-->>PluginMain: Return mapped full-body frame
  PluginMain->>FullBodyPublisher: Publish frame
  FullBodyPublisher->>OpenXRTransport: Push serialized pose
Loading
Merge Risk: Minimal · up to a91f3

This change adds an optional j-mex full-body plugin, a body.jmex tracker, and examples. No concrete defects are established, and builds without the SDK skip the plugin, so the change appears ready to merge.

Pre-merge checks: Passed 4

Check name Status Explanation
Linked Issues check Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check Passed Check skipped because no linked issues were found for this pull request.
Title check Passed The title clearly and concisely describes the main changes: adding the AgileMaster full-body plugin and the G1 teleoperation example.
Description Check Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Finishing Touches

🧪 Generate unit tests (beta)

  • Create a new PR
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
ℹ️ Recent review info

⚙️ Run configuration

  • Configuration used: Repository: NVIDIA/IsaacCapture/.coderabbit.yaml
  • Review profile: CHILL
  • Plan: Enterprise
  • Run ID: b150d31f-05b8-4673-9d90-291312365719
📥 Commits

Reviewing files that changed from the base of the PR and between 2fe1bbe and a91f3af.

📒 Files selected for processing (41)

  • cmake/ClangFormat.cmake
  • docs/source/_data/devices.yaml
  • examples/jmex/.gitignore
  • examples/jmex/README.md
  • examples/jmex/pyproject.toml
  • examples/jmex/python/isaaccapture_examples/jmex/__init__.py
  • examples/jmex/python/isaaccapture_examples/jmex/jmex_g1_retargeting.py
  • examples/jmex/python/isaaccapture_examples/jmex/jmex_g1_tasks.py
  • examples/jmex/python/isaaccapture_examples/jmex/record_jmex_full_body.py
  • src/core/live_trackers/cpp/CMakeLists.txt
  • src/core/live_trackers/cpp/inc/live_trackers/live_deviceio_factory.hpp
  • src/core/live_trackers/cpp/live_deviceio_factory.cpp
  • src/core/live_trackers/cpp/live_full_body_tracker_jmex_impl.cpp
  • src/core/live_trackers/cpp/live_full_body_tracker_jmex_impl.hpp
  • src/plugins/CMakeLists.txt
  • src/plugins/jmex/.gitignore
  • src/plugins/jmex/CMakeLists.txt
  • src/plugins/jmex/README.md
  • src/plugins/jmex/app/CMakeLists.txt
  • src/plugins/jmex/app/main.cpp
  • src/plugins/jmex/core/CMakeLists.txt
  • src/plugins/jmex/core/body_joint_mapping.cpp
  • src/plugins/jmex/core/device_clock.cpp
  • src/plugins/jmex/core/full_body_publisher.cpp
  • src/plugins/jmex/core/inc/jmex/body_joint_mapping.hpp
  • src/plugins/jmex/core/inc/jmex/device_clock.hpp
  • src/plugins/jmex/core/inc/jmex/full_body_publisher.hpp
  • src/plugins/jmex/core/inc/jmex/receiver_session.hpp
  • src/plugins/jmex/core/openxr_convention.cpp
  • src/plugins/jmex/core/openxr_convention.h
  • src/plugins/jmex/core/receiver_session.cpp
  • src/plugins/jmex/plugin.yaml
  • tests/cpp/core/live_trackers/test_vendor_validation.cpp
  • tests/cpp/plugins/CMakeLists.txt
  • tests/cpp/plugins/jmex/CMakeLists.txt
  • tests/cpp/plugins/jmex/test_jmex_plugin_core.cpp
  • tests/python/examples/CMakeLists.txt
  • tests/python/examples/jmex/CMakeLists.txt
  • tests/python/examples/jmex/conftest.py
  • tests/python/examples/jmex/pyproject.toml
  • tests/python/examples/jmex/test_jmex_g1_action_source.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

@ivany-nv ivany-nv self-assigned this Oct 13, 2026
@ivany-nv
ivany-nv self-requested a review October 13, 2026 18:16
@ivany-nv

Copy link
Copy Markdown
Contributor

/preview-docs

@github-actions

Copy link
Copy Markdown
Contributor

✅ Preview deployed: https://NVIDIA.github.io/IsaacCapture/preview/pr-1250/

@ivany-nv

Copy link
Copy Markdown
Contributor

@aristarkhovNV Can you please help review the core change part.

@ivany-nv ivany-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Summary
This adds a MOXI Receiver plugin that publishes AgileMaster poses as FullBodyPose, a body.jmex live tracker that reads them (same shape as the Noitom and Xsens readers), and an Isaac Lab G1 example. CI is green on every job. Built locally without the SDK, the jmex tests pass, and the plugin's own publisher, run against a CloudXR runtime, round-trips through body.jmex into a standard full-body MCAP with every joint value intact. Read independently, the axis change, slot mapping, quaternion order, timestamps and MCAP wiring all hold up. One suggestion, nothing blocking. The src/core/live_trackers reader and factory registration are with @aristarkhovNV; this review covers the plugin, example, tests and docs.

Legend: 🚫 BLOCKER = merge-blocking | 💡 SUGGESTION = meaningful improvement | 🧹 NIT = precise small correction

Finding
🚫 none
💡 src/plugins/jmex/app/main.cpp:126-129 - An unrecognized argument is only logged as a warning. So --channel 2 (space instead of =) or a typo like --chanel=2 leaves the plugin on channel 1, waiting for Player forever. That is the silent mismatch the strict --channel= parsing at L92-L93 exists to prevent. The Xsens plugin rejects unknown options; doing the same here makes the plugin fail fast with exit code 1.
🧹 none

@ivany-nv ivany-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Summary
A second pass, this time against this PR merged onto current main, which is 55 commits past your base. It merges cleanly and builds. ctest -L jmex, the vendor-validation cases, clang_format, pre-commit on the changed files and make current-docs all pass. The body.jmex reader is safe to add: the default vendor is unchanged, there is no schema or recording change, and it compiles against main's newer SchemaTracker. Before approval there are three things: two example fixes (the first two 💡 below) and a rebase onto main so CI reruns against it. The rest are questions and small items.

Legend: 🚫 BLOCKER = merge-blocking | 💡 SUGGESTION = meaningful improvement | ❓ QUESTION = needs your answer | 🧹 NIT = precise small correction

Finding
🚫 none
💡 jmex_g1_tasks.py:30-32 - Isaac Lab 3.0.0 EA, the release main's README names as current, moved this config to isaaclab_tasks.contrib.locomanip_pick_place.locomanipulation_g1_env_cfg (isaac-sim/IsaacLab#5891). The manager_based path no longer exists there, so --external_callback ...register_tasks fails with ModuleNotFoundError. Please import the new path with a fallback to the beta2 one, and name the tested Isaac Lab version in README prerequisite 4.
💡 jmex_g1_tasks.py:711-725 - Returning [] tells teleop_se3_agent.py to reject none of the leftover arguments: list_intersection(remaining_args, []) is empty, and with None it is all of them. So --cloudxr-env none (hyphen) is silently ignored and Isaac Lab starts its own auto-webrtc runtime. That is the failure README L97-99 warns about, now with no error. The callback parses no arguments of its own, so return None is right.
💡 main.cpp:59-75 - This extends my earlier comment. parse_port checks that the whole string is an integer but not its range. So --tcp-port=70000, --channel=0 or --channel=300 go straight to the SDK, and udp_start_port + 1 + channel (L151) can overflow. The noitom and xsens plugins reject out-of-range ports, and receiver_session.hpp:27-28 already states the valid channels.
❓ app/CMakeLists.txt:8-14 - Is Moxi::Receiver static or shared? If it is shared, the installed jmex_plugin has no RPATH to find it. The plugins that link a vendor shared library set INSTALL_RPATH; sharpa_avatar points it at an external SDK. readelf -d install/plugins/jmex/jmex_plugin on your machine answers this.
❓ receiver_session.cpp:136-153 - The joint table is re-read only after the channel drops. Can Player load a different model, as the README's troubleshooting tells operators to do, without the pairing dropping? If it can, slots keep the old joint indices and go out valid. Re-checking MxRGetBoneJointCount on each new sequence id would cover it.
❓ jmex_g1_tasks.py:769-777 - retargeting_execution is unset, so Isaac Lab runs the pipeline in pipelined mode, and the overlay's debug-draw calls run on the retarget worker thread. Which mode was the live run in? Setting mode="sync", or drawing from process_actions, settles it either way.
🧹 .gitignore:5 / ClangFormat.cmake:43 - Use MOXIReceiverSDK*/ and MOXIReceiverSDK[^/]*, so an archive unpacked under its versioned name is not picked up by git add -A or the format check.
🧹 full_body_publisher.hpp:42-48 - body.jmex accepts max_flatbuffer_size, but the plugin always creates a 16 KiB collection, so a smaller value makes the reader throw. Either drop the parameter or add --max-flatbuffer-size= as noitom and xsens do.
🧹 device_clock.cpp:52-60 - A Player timestamp of 2^63 or more turns negative, push_buffer throws, and publish() treats every such frame as a lost session. Falling back to the sample time there keeps push failures transport-only.
🧹 Naming, after the rebase: main now calls the library and the repository "Isaac Capture"; Isaac Teleop stays the teleop component. This affects the plugin README L8-9, the example README L11/29/31/49 and record_jmex_full_body.py:4. The devices.yaml row's URL should be NVIDIA/IsaacCapture. The example README's wheel steps (L22-43) can link the build-from-source page and keep only the Isaac-Lab-specific parts.
🧹 src/plugins/jmex/CMakeLists.txt - The README says Linux x86_64 only. A CMAKE_SYSTEM_NAME check with a "Linux only" skip message, as the xsens and noitom plugins have, reports the right reason on other platforms.

Optional, not needed for approval:

  • Build the SDK-free sources as an always-built core library, with the receiver in the SDK-gated part. The test could then link it, and those files would get the project warning set, as gamepad and spacemouse do.
  • Split JmexG1ActionSource into FullBodySource plus a retargeter, to match the IDeviceIOSource "pure converter" contract.

Please rebase onto main, or we can update the branch, so CI reruns against it before approval.

Adds `jmex`, a plugin for the j-mex AgileMaster motion capture suit, and an example, in two parts,
that consumes what it publishes.

The plugin reads one MOXI Player channel through the MOXI Receiver SDK (0.8.0 or newer, obtained
from j-mex) and runs on Linux only. It fills the 24 `FullBodyPose` slots from the standard human
skeleton Player has loaded, one slot per joint, already in OpenXR axes, and pushes them under the
`body.jmex` vendor on every frame. A slot with no joint, or with no usable value in a frame, goes
out invalid; nothing is inferred from its neighbors. The raw device clock carries the time Player
stamped on the frame (`--device-clock=local` uses the pickup time instead), and if the CloudXR
runtime restarts underneath the plugin, the plugin reconnects rather than ending the run. An
argument it cannot use stops it with exit code 1, naming the argument.

Everything but the receiver is a library that builds without the SDK, so on Linux its unit tests
run in every build that has tests enabled; without the SDK only the plugin executable is skipped.

Isaac Capture learns `body.jmex` as a full-body vendor: a live tracker that reads the plugin's
collection, with a `collection_id` parameter, its entry in the DeviceIO factory, the vendor
validation cases, and a row in `devices.yaml`. An SDK unpacked into the plugin directory, under any
versioned name, is kept out of git and the clang-format check.

`isaaccapture_examples.jmex.jmex_g1_tasks` registers a task that inherits Isaac Lab's
`LocomanipulationG1EnvCfg`, on Isaac Lab 3.0.0 EA or 3.0.0 beta2, and takes its action from the
operator; nothing in Isaac Lab is modified. It is upper-body teleop. Each arm is rebuilt segment by
segment onto the robot's own shoulder, so the operator's arm angles survive while the lengths stay
the robot's, and the elbows and shoulders get IK targets of their own from that same chain. The
wrist's orientation is the operator's hand, carried across through a forearm frame. The operator's
torso turns the robot's shoulders about its waist, so leaning, bending sideways and twisting come
through; their heading does not, so the operator can turn to face a screen without moving the
robot. Only a frame that carries all 24 slots drives the robot, and slots that stay missing are
named on the console.

`isaaccapture_examples.jmex.record_jmex_full_body` writes the same stream to a standard full-body
MCAP.

Tests cover the plugin's argument parsing, slot mapping, change of axes, device clock choice and
session rebuild after a failed push, the vendor's parameters, and the example's action source,
including the geometry of the rebuilt arm and the orientation carried through the forearm frame.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: j-mex <jmex.dev@jmex.com.tw>
@jmex-sw
jmex-sw force-pushed the jmex-plugin-full-body branch from a91f3af to 8dcc95c Compare October 30, 2026 09:26
@jmex-sw

jmex-sw commented Oct 30, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the second pass. I rebased onto main at 49a218122 and force-pushed: the branch is
still one commit, now 8dcc95c53, and everything below is in it.

Finding Done
💡 Isaac Lab 3.0.0 EA path jmex_g1_tasks.py imports the isaaclab_tasks.contrib.locomanip_pick_place config and falls back to the beta2 path only when that module is missing. README prerequisite 4 and "Verified on" name v3.0.0-EA (Isaac Sim 6.1.0) and v3.0.0-beta2.patch1 (Isaac Sim 6.0.1). Running on EA also showed that the overlay enabled its debug-draw extension through isaacsim.core.experimental, which EA's app does not load; it now goes through Kit's extension manager. The README no longer uses --headless, and notes that on EA Ctrl-C leaves Isaac Lab running, as it does for EA's own tasks.
💡 register_tasks() returning [] Returns None.
💡 Argument ranges Parsing moved into the core library as parse_plugin_args, with unit tests. A channel outside 1–255 (the SDK keeps the channel in one byte; Player uses 1–10 live and 255 for replay), a port outside 1–65535, or a channel data port past 65535 stops the plugin with exit code 1, naming the argument. So does an unknown or positional argument, which covers your first-round comment too.
❓ Static or shared Static: the SDK ships only libmoxi_receiver.a. readelf -d on the installed jmex_plugin shows no RPATH or RUNPATH, and its only NEEDED entry besides the C and C++ runtime is libjsoncpp.so.25, which the SDK does not use and most of the in-tree plugins link as well.
❓ Model change without a disconnect Player always drops the pairing before it loads another model, and the plugin re-reads the joint table on the next pairing. Checked with Player on channel 255: standard human skeleton, 65 joints, 24 of 24 slots; loading a robot model logged MOXI Player disconnected, then paired, 50 joints with 4 of 24 slots and the missing ones named; loading the skeleton again logged the disconnect, then paired, 65 joints, 24 of 24.
❓ Retargeting mode The live run was pipelined, Isaac Lab's default. The task now sets RetargetingExecutionConfig(mode="sync"). Headless, the step rate is unchanged (about 11 steps/s on EA, 12 on beta2), and the overlay was checked in Kit on EA. Remeasured live on beta2 in sync: the waist follows at the same 90–270 ms, with correlation 0.97–0.99, as the PR description says.
🧹 Versioned SDK directory MOXIReceiverSDK*/ in .gitignore, MOXIReceiverSDK[^/]* in the format check's exclusions.
🧹 max_flatbuffer_size Dropped: body.jmex takes only collection_id, and both ends use 16 KiB. This is the only change under src/core since the first review.
🧹 Player time past int64 Falls back to the sample time, with a test at the boundary.
🧹 Naming Isaac Capture where the library or the repository is meant, the devices.yaml URL, and the example's wheel step now links Build from source and keeps only what differs for Isaac Lab.
🧹 Linux only CMAKE_SYSTEM_NAME check with a "Linux only" skip message; the plugin's tests moved into the Linux-only block.
Optional Core library Done. Everything but the receiver is jmex_plugin_core, built with or without the SDK and with the project warning set, and the tests link it. The receiver is compiled into jmex_plugin in app/, the only part that needs the SDK.
Optional Splitting JmexG1ActionSource Not in this PR. I'd like to do it as a follow-up.

The PR description has the updated local gate (both toolchains, six entries each) and the device
checks. The same commit passed the Build Ubuntu workflow on our fork: https://github.com/jmex-sw/IsaacTeleop/actions/runs/41028280279. Could you approve
the workflow runs here when you have a moment?

@ivany-nv

Copy link
Copy Markdown
Contributor

Heads-up: #1394 just merged. It moves only-include from [tool.hatch.build.targets.wheel] to [tool.hatch.build], so the sdist gets the same file list as the wheel. Without that, uv build packs the whole example directory into the sdist: media, recordings, a local uv.lock, build trees. Wheels and editable installs are unchanged.

Could you switch examples/jmex/pyproject.toml to the layout in Adding an example?

[tool.hatch.build]
only-include = ["python/isaaccapture_examples/jmex"]

[tool.hatch.build.targets.wheel]
sources = ["python"]

Nothing checks where only-include lives, so rebasing won't flag the old layout.

main conflicted in tests/cpp/plugins/CMakeLists.txt and now asks
examples to set only-include for the sdist too. On top of the merge:

- tests/cpp/plugins/CMakeLists.txt registers the plugin's tests under
  if(TARGET jmex_plugin_core) at the end of the file instead of in the
  Linux-only block, so no existing line of the file changes. The core
  library exists on Linux, with or without the SDK.
- examples/jmex/pyproject.toml sets only-include in [tool.hatch.build],
  so the sdist gets the same file list as the wheel.
- The example's recording and replay steps and the plugin's runtime
  step start from a venv set up as in Running an example.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: j-mex <jmex.dev@jmex.com.tw>
@jmex-sw

jmex-sw commented Nov 3, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the heads-up. I merged main at 834ecad34 into the branch, so the PR head is now
3f3bb7be6, one merge commit on top of 8dcc95c53; nothing was force-pushed.

  • examples/jmex/pyproject.toml sets only-include in [tool.hatch.build], as in Adding an example.
    Built from a copy with media, a stray recording, a uv.lock and a build tree beside the package,
    the sdist holds only the package, pyproject.toml, .gitignore and PKG-INFO, and the wheel built
    from the unpacked sdist has the same four modules as before.
  • main conflicted in tests/cpp/plugins/CMakeLists.txt, next to the Linux-only note this PR had
    edited. The plugin's tests now register under if(TARGET jmex_plugin_core) at the end of that
    file, so the PR no longer changes any existing line there. The core library is defined on Linux,
    with or without the SDK, as before.
  • The example's recording and replay steps and the plugin README's runtime step now start from a venv
    set up as in Running an example.

The PR description has the local gate rerun on this tree. The same commit passed the Build Ubuntu
workflow on our fork: https://github.com/jmex-sw/IsaacTeleop/actions/runs/41576012036. Could you
approve the workflow runs for the updated head 3f3bb7be6?

@github-actions

github-actions Bot commented Nov 4, 2026 •

Copy link
Copy Markdown
Contributor

✅ Preview deployed: https://NVIDIA.github.io/IsaacCapture/preview/pr-1250/

Built from commit fa1b5e0. It may take a minute for the updated preview to appear.

@ivany-nv

ivany-nv commented Nov 4, 2026

Copy link
Copy Markdown
Contributor

Review in progress.

@ivany-nv ivany-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Summary
Thanks. Every item from the last round is fixed. Merged onto main (75d567089), the jmex code builds without warnings, all jmex tests (23 Catch2, 91 pytest) pass, and upstream CI is green. What is left is inline: 1 🚫, 3 💡, 1 ❓, 8 🧹.

Legend: 🚫 BLOCKER = merge-blocking | 💡 SUGGESTION = meaningful improvement | ❓ QUESTION = needs your answer | 🧹 NIT = precise small correction

The CTS review has passed.

Comment thread src/plugins/CMakeLists.txt Resolved
Comment thread src/plugins/jmex/plugin.yaml Resolved
Comment thread src/plugins/jmex/core/plugin_args.cpp Resolved
Comment thread src/plugins/jmex/README.md Resolved
Comment thread examples/jmex/README.md Resolved
Comment thread docs/source/_data/devices.yaml Resolved
Comment thread src/plugins/jmex/app/main.cpp Resolved
Comment thread src/plugins/jmex/core/plugin_args.cpp Resolved
main conflicted in tests/cpp/plugins/CMakeLists.txt and
tests/python/examples/CMakeLists.txt, where it added a subdirectory
next to jmex's; both are kept. On top of the merge:

- BUILD_PLUGIN_JMEX, default OFF, gates src/plugins/jmex and the
  example's Python tests; the C++ tests follow through
  if(TARGET jmex_plugin_core). With the option on and no SDK, configure
  warns and names it. The READMEs' build commands and the example's
  install hints pass -DBUILD_PLUGIN_JMEX=ON.
- The plugin needs MOXI Receiver SDK 0.8.1, which advertises only from
  the TCP server's address when one is given. plugin.yaml binds the
  receiver to loopback, so a plugin TeleopSession launches pairs only
  with a Player on this machine; the README says what a Player on
  another machine needs, and that the channel's data port takes frames
  from any sender. The plugin no longer checks the data port's range
  itself: 0.8.1's MxROpenChannel fails on it, and the plugin says so.
- --help or -h prints the flags and their defaults and exits 0. An
  empty --collection-id=, --bind-ip= or --broadcast-ip= stops the
  plugin. --channel=0, which Player uses for a suit on USB, is
  accepted.
- The G1 task pins the root link on the Isaac Lab v3.0.0 release, where
  G1_29DOF_CFG's articulation_props is a list, and the example README
  lists v3.0.0 under Verified on.
- Comments and wording: why the task retargets in sync, the base config
  named without a beta2-only path, the docs-site build link, MOXI
  Player for Robot, and Linux x86_64 in the device row. A test no
  longer claims its model is a G1 URDF, and the data-port case has a
  test of its own.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: j-mex <jmex.dev@jmex.com.tw>
@jmex-sw

jmex-sw commented Nov 9, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the review, and for the CTS result. The PR head is now d08d4b73d, one merge commit on top
of 3f3bb7be6 with main at 4c7d4db9a; nothing was force-pushed.

Finding Done
🚫 BUILD_PLUGIN_JMEX Added after the other device plugins' options, default OFF. src/plugins/jmex and the example's Python tests are entered only under it; the C++ tests follow through if(TARGET jmex_plugin_core). Both READMEs' build commands and the install hints in jmex_g1_tasks.py and record_jmex_full_body.py pass -DBUILD_PLUGIN_JMEX=ON. With the option on and no SDK, configure warns BUILD_PLUGIN_JMEX=ON but no MOXIReceiverSDK 0.8.1 or newer was found. A default configure from an empty tree, with CI's toolchain, has no jmex target, build directory or test.
💡 Isaac Lab v3.0.0 fix_root_link goes on articulation_props when that config has the field (3.0.0 beta2 and EA) and on spawn otherwise. On v3.0.0 with Isaac Sim 6.2.0 the README's steps ran headless: the task launched the plugin, a replayed Player capture paired, and each of 900 steps was computed from a full body pose. "Verified on" lists it.
💡 Loopback in plugin.yaml With SDK 0.8.0, Player could not pair over loopback: the SDK advertised from every non-loopback interface whatever the TCP address was, Player connected back to the interface address, and a TCP server on 127.0.0.1 refused it. MOXI Receiver SDK 0.8.1 advertises only from the TCP server's address when one is given, so the plugin now requires 0.8.1, and plugin.yaml passes --bind-ip=127.0.0.1 --broadcast-ip=127.0.0.1 with a comment. With Player on the same machine it paired within 6 s, with the TCP server listening only on 127.0.0.1. For a Player on another machine, the README says to set --bind-ip= to this machine's address facing it but --broadcast-ip= to Player's address, not to the local NIC: advertisements go to that address and Player connects back to their source, so advertising to the local NIC would reach only this machine. Plus a firewall rule for the TCP port and the data ports. With Player for Robot on a second machine on the same subnet, those two settings paired within 5 s in both runs; with --broadcast-ip= set to the local NIC, or with plugin.yaml's loopback flags, Player did not connect in 40 s.
💡 Empty --collection-id= Stops the plugin, with a test. So do an empty --bind-ip= and --broadcast-ip=, which would otherwise fall back to every interface and 255.255.255.255.
❓ Data port and sender No. The SDK binds the data port on every interface, whatever --bind-ip= says, and does not check the sender, so a datagram from any host reaches a paired channel. The README says so beside the network defaults: only a firewall keeps other hosts out, and --udp-start-port= makes the data ports predictable enough for a rule.
🧹 Sync comment # Required: pipelined, the default, would run the overlay's Kit calls on a worker thread.
🧹 Path that exists only in beta2 The comment names the pipeline Isaac Lab builds for LocomanipulationG1EnvCfg, without a path.
🧹 Build from Source link Now the docs-site page.
🧹 Second plugin The sentence is gone; the flag table says what each plugin needs.
🧹 MOXI Player for Robot (example README) Done.
🧹 devices.yaml MOXI Player for Robot, and Linux, x86_64 as a requirement.
🧹 Data-port formula The plugin no longer works the port out. The SDK picks it, moves it when it is taken and tells Player, and has no call that reports it, so the startup log prints the base it was given. The range check is gone too: SDK 0.8.1's MxROpenChannel fails when it can bind no data port, past 65535 included, and the plugin then stops and says so.
🧹 --help --help or -h prints the flags with their defaults and exits 0, before anything else is read.

One change you did not ask for: --channel=0 is now accepted. MOXI Player streams a suit connected
over USB on channel 0, and the plugin used to stop on it.

The PR description has the local gate rerun with -DBUILD_PLUGIN_JMEX=ON, plus one entry configured
as CI does. The same commit passed the Build Ubuntu workflow on our fork: https://github.com/jmex-sw/IsaacTeleop/actions/runs/42554595102. With the
option off by default, those jobs build nothing of the plugin. Could you approve the workflow runs
for the updated head d08d4b73d?

Comment thread docs/source/_data/devices.yaml Resolved
Comment thread src/plugins/jmex/core/plugin_args.cpp Resolved
Comment thread tests/cpp/plugins/CMakeLists.txt Resolved
main merged without conflicts. On top of the merge:

- The G1 example reads the suit through FullBodySource with the
  body.jmex vendor. JmexG1Retargeter, a BaseRetargeter on the optional
  full_body input, holds the state kept across frames and outputs the
  frame tasks' poses. LocomotionFixedRootCmdRetargeter gives the
  standing command and TensorReorderer lays out the action, with the
  Pink IK action term's hand_joint_dim fingers left at zero.
  JmexG1ActionSource, the module-level runtime and the example's own
  action packing are gone.
- The example's plugin name, ids, defaults, search path and launcher
  live in one module, plugin.py, which looks only under
  install/plugins. Imports are at the top of each module, except Kit's
  two, which exist only inside a running Kit app, and Isaac Lab's Pink
  IK action class, which Isaac Lab's own config names by a string. The
  example's loggers are isaaccapture.examples.jmex.*.
- The plugin takes its settings from flags only; JMEX_DEVICE_CLOCK is
  gone, and plugin.yaml no longer repeats the default channel and
  collection. parse_plugin_args returns std::optional<PluginArgs>,
  empty for --help. A scope object declared once the receiver is up
  logs the counters, however the run ends.
- ReceiverSession keeps each pairing's state in one
  std::optional<Pairing>, and holds the SDK's system and channel as
  RAII members; the process-wide session claim is gone.
  BodyJointMapping is built by its constructor and takes core::Pose.
  Impossible inputs are asserted, and optional addresses and joint
  indices are std::optional.
- CMake finds the SDK through CMAKE_PREFIX_PATH, MOXIReceiverSDK_DIR or
  the plugin directory; JMEX_SDK_ROOT is gone. On a target other than
  x86_64, configure warns and builds only the core library and its
  tests. jmex_plugin_core drops its PIC and visibility properties.
- The docs no longer state the SDK version or the recovery time. The
  tests no longer sleep or check constant values, the Catch2 tests
  carry a jmex_ prefix, and a duplicate body.jmex vendor-validation
  section is gone. The core library's alias is jmex::jmex_plugin_core.
  A joint-mapping test feeds joints out of slot order.
- A lower sequence id after UINT32_MAX no longer counts as dropped
  frames, the recording script times itself with time.monotonic(),
  and FrameTargetCfg refuses non-finite link lengths and quaternions.
- Numeric flags are parsed with std::from_chars, so a leading space or
  plus sign is refused, and the plugin's CMakeLists.txt no longer
  repeats cmake_minimum_required.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: j-mex <jmex.dev@jmex.com.tw>
@jmex-sw

jmex-sw commented Nov 12, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the review. The PR head is now 550257dbd, one merge commit on top of d08d4b73d with
main at 16578fe0e; nothing was force-pushed.

# Comment Response
1 SDK version in three places devices.yaml and the plugin README no longer name a version. It is written once, in src/plugins/jmex/CMakeLists.txt, and configure's warning prints it when no SDK that new is found.
2 num_locomotion Deleted, with the layout's locomotion_command and num_hand_joints fields: the example no longer builds that part of the action (#3).
3 FullBodySource Done. The pipeline is FullBodySource (vendor body.jmex) → JmexG1Retargeter, a BaseRetargeter on the optional full_body input; an absent input takes the unpaired path. The held chains and orientations, the operator measurement and the report counters are the retargeter's state, and the change into Isaac axes happens there. The retargeter outputs only the frame tasks' poses: LocomotionFixedRootCmdRetargeter, with its default config, gives the standing command, and TensorReorderer lays out the action with the Pink IK action term's hand_joint_dim fingers left at zero, the way Isaac Lab composes its own G1 locomanipulation pipeline. JmexG1ActionSource is gone. The tests drive BaseRetargeter.compute with FullBodyInput tensor groups or an absent group, and one test checks the composed action's length and order without Isaac Lab.
4 layout Removed, and so are settings and operator_segment_lengths, which only the tests read. The operator measurement is tested on its log line.
5 Missing .joints Gone with the old source node: the retargeter reads FullBodyInput's arrays as they arrive, with no zero fill.
6 _env can be None No longer: the runtime takes the env and the layout together and raises if either is read before that. The zero origin is gone.
7 Module-level singleton Gone. __post_init__ builds one _JmexRuntime and hands it to the action class factory, the pipeline builder, the overlay drawer and the action-length check, which all close over it. It is not a cfg field, so a copied configclass keeps them on the same instance.
8 <root>/plugins It never exists for this example. Only install/plugins is searched now.
9 Imports pinocchio and pink.tasks are at the top. Three imports stay inside functions. omni.kit.app and isaacsim.util.debug_draw exist only inside a running Kit app, and the second only after the overlay enables its extension, one line earlier. PinkInverseKinematicsAction is imported inside the factory that subclasses it, so importing the task module does not load Isaac Lab's action stack. The repository's other Isaac Lab G1 example builds its Pink IK action subclass the same way, and Isaac Lab's own PinkInverseKinematicsActionCfg names the class by a string instead of importing it.
10 Redundant check Removed, with the "as in the tests" note.
11 Duplicated constants Defined once, in a new isaaccapture_examples.jmex.plugin module: plugin name, root id, vendor id, and the default collection and channel. JmexG1Settings, the task and the recording script take them from there. The module also builds the PluginConfig and the FullBodySource the task and the script both use.
12 _plugin_search_paths One function, plugin.plugin_search_paths().
13 Flags in the --no-plugin help Removed. The help points to the plugin README and jmex_plugin --help.
14 JMEX_SDK_ROOT Removed: the cache variable, the environment variable and the cache reset. One find_package searches CMAKE_PREFIX_PATH, MOXIReceiverSDK_DIR and the plugin's MOXIReceiverSDK/ directory. The warning and the README name the two standard variables.
15 FATAL_ERROR for an explicit root With JMEX_SDK_ROOT gone, the plugin has no variable of its own left to check. We measured what CMake 3.28 does with a MOXIReceiverSDK_DIR that holds no SDK. Set alone, configure prints the same warning as with no SDK and skips the plugin. Set together with a CMAKE_PREFIX_PATH that holds one, find_package uses that SDK without a warning; the status line jmex plugin: MOXIReceiverSDK <version> from <dir> names the directory it used. The same holds for an SDK unpacked into the plugin's MOXIReceiverSDK/ directory: with one there, a wrong MOXIReceiverSDK_DIR builds against it without a warning. If you want configure to stop in that case, we can add the check.
16 23 s Both places describe the behaviour instead: the plugin retries with a growing backoff, and a runtime still gone when the retries run out ends the run with exit code 1.
17 Is Moxi::Receiver static? Yes. The SDK's MOXIReceiverSDKTargets.cmake declares it STATIC IMPORTED on libmoxi_receiver.a, and ldd on the installed jmex_plugin lists no SDK library. A one-line comment at the link line now says so.
18 JMEX_DEVICE_CLOCK None that we know of. The variable is gone, with the precedence rules and their tests; --device-clock= is the one way to set it.
19 Inner try-catch Gone. CountersOnExit, declared after receiver, logs the counters from its destructor, and the exception goes on to main's handler. A clean stop still ends on the counters line; on an error, the counters line now comes before the error. A failure before the receiver is up logs only the error.
20 Hardcoded 24 core::BodyJoint_NUM_JOINTS.
21 Does the SDK give a reason? No. MxRStartSystem and MxROpenChannel return only a bool, and MxRBindDebugLogCallback delivers messages only from an SDK built with logging on. Both messages now say the call returned false and list every cause MoxiReceiver.h documents for it, without singling one out.
22 Manual unwind The system and the channel are each a small RAII member, so the constructor does no cleanup and ReceiverSession has no destructor of its own. g_session_alive is deleted rather than moved into the system member: MxRStartSystem already returns false while a receiver is running, which its error message lists, and main builds one session.
23 0, -1 and empty strings as sentinels "No frame yet" is an empty std::optional<Pairing>; the last sequence id inside it is always set. The SDK's own 0, which it returns before a channel's first frame, is checked once, where the id is read. BodyJointMapping::joint_index() returns std::optional<size_t>, and bind_ip and broadcast_ip are std::optional<std::string>.
24 Per-pairing state One std::optional<Pairing> holds the last sequence id and the mapping, the mapping itself optional until the joint table arrives. It is reset when the channel stops streaming. joints_discovered_ and the joint name member are gone. was_streaming_ stays beside it: motion that arrives before a reference pose sets the channel streaming without moving its sequence id, so streaming and paired can differ. Its comment says so.
25 PIC and visibility Removed.
26 Slot range check assert.
27 Short pose vector assert(joint_poses.size() == joint_count()), since ReceiverSession sizes the vector from the mapping's joint_count(). The promise that an empty frame invalidates every slot is gone, and so is its test.
28 Null transport The constructors bring a transport up or throw, so a publisher never starts without one. The one gap is inside a recovery: the old transport is released before the new one is requested, so two sessions never exist at once. A recovery ends with a transport, a throw, or stop set. publish() now asserts the transport instead of checking it.
29 core::Pose Done. JointPose and to_schema() are gone; moxi_pose_to_openxr, apply and is_usable use core::Pose.
30 resolve() as constructor Done. The explicit constructor builds the mapping from the joint table, and there is no default constructor; ReceiverSession holds the mapping as an optional inside Pairing.
31 help == true parse_plugin_args returns std::optional<PluginArgs>, empty for --help or -h. It prints nothing; main prints the usage once and exits 0, before it creates its logger. Errors still throw and exit 1.
32 starts_with The helper is gone. In the parser, each flag's name is now written once, matched by a flag_value() that uses std::string_view::starts_with, as the gamepad plugin's does.
33 Tests never run in CI Right: CI leaves BUILD_PLUGIN_JMEX off, so they do not run there today. Not changed yet; see below.
34 Real sleeps That section is gone, and so is Threads. The default backoff wait now has no unit test.
35 Testing constants The usage tests no longer check default values; the one that checks --help names every flag stays. STANDING_LOCOMOTION_COMMAND went with #2 and #3, its asserts with it.

On #33. Round 4 asked for the opposite. In round 2 an optional suggestion was to build
the SDK-free part as a core library that always builds, and we did that. Round 4 then blocked it:

This line is unconditional, so every default configure, Build Ubuntu included, builds
jmex_plugin_core and runs its tests. Every other partner plugin option defaults to OFF
(CMakeLists.txt:70-73); whether a partner plugin runs in CI is for the CI owners to decide.

It also asked for tests/python/examples/CMakeLists.txt to be gated the same way, which is how the
PR stands now. Either way is a small change. Which would you both prefer? We will follow it.
With the option on and no SDK, configure already builds jmex_plugin_core and both test suites, so
CI entries that turn the option on would run them without the SDK.

Changes you did not ask for:

  • On a target other than x86_64, configure warns and skips the plugin executable: the SDK ships
    for x86_64 only. The core library and its tests still build.
  • A new sequence id is compared with the last one, not with the last one plus one, which wraps
    when the last id is UINT32_MAX; a lower id after it no longer counts as dropped frames.
    ReceiverSession links the SDK, so this has no unit test.
  • FrameTargetCfg refuses NaN and infinite link lengths and quaternions at construction; the
    existing <= 0 and norm checks let them through.
  • The recording script measures its duration and progress with time.monotonic().
  • Numeric flags are parsed with std::from_chars, as core's vendor parameters are, so --channel=+5
    and --channel= 5 are refused.
  • The Catch2 tests carry TEST_PREFIX "jmex_", so ctest -R jmex_ selects them.
  • The library alias is jmex::jmex_plugin_core, the mod::mod form of cmake/cmake-structure.md.
  • The example's loggers are isaaccapture.examples.jmex.*, under the examples' namespace.
  • In test_vendor_validation.cpp, the body.jmex section for an unknown parameter is gone: the
    max_flatbuffer_size section checks the same rejection.
  • plugin.yaml no longer passes --channel= and --collection-id=; they equalled the plugin's
    defaults.

The PR description has the local gate rerun on this tree. The
same commit passed the Build Ubuntu workflow on our fork: https://github.com/jmex-sw/IsaacTeleop/actions/runs/43162912770. With this round's
pipeline, the task ran headless on Isaac Lab 3.0.0 beta2, 3.0.0 EA and v3.0.0 from a replayed
Player stream, 900 steps each. On v3.0.0, with Kit's window and MOXI Player on the same machine,
the G1 followed, held its pose while Player was disconnected, and followed again once Player
paired again. Could you approve the workflow runs for 550257dbd?

@ivany-nv

Copy link
Copy Markdown
Contributor

@jmex-sw Everything looks good, please address the last 2 left item. See comment under AristarkhovNV's comment. Thanks

…t the option

- A MOXIReceiverSDK_DIR that is set is the only place configure looks
  for the SDK. If it holds no MOXIReceiverSDK of the minimum version
  or newer, configure stops with a FATAL_ERROR that names the variable
  and its value, rather than building against an SDK found through
  CMAKE_PREFIX_PATH or in the plugin directory. find_package rewrites
  a value it cannot use to -NOTFOUND in the cache, so the value is
  written back first and the next configure stops too. Without the
  variable, the search and the warning are as before.
- The plugin's SDK-free core library and its Catch2 tests build on
  Linux whether or not BUILD_PLUGIN_JMEX is on, and the example's
  Python tests are no longer gated on it. The option, still OFF by
  default, adds only the jmex_plugin executable and its install.
- The plugin README's Building section and Troubleshooting table cover
  the new error, and the example README's offline-test section no
  longer asks for the option.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: j-mex <jmex.dev@jmex.com.tw>
@jmex-sw

jmex-sw commented Nov 17, 2026

Copy link
Copy Markdown
Contributor Author

Thanks. The PR head is now fa1b5e067, one commit on top of 550257dbd; nothing was force-pushed.
main was not merged this time, since the PR has no conflicts with it. Both items are done, with a
reply under each.

The PR description's Testing section has the local gate rerun on this tree. Every matrix entry left
BUILD_PLUGIN_JMEX off, as CI does, and ran the 25 jmex tests. The same commit passed the
Build Ubuntu workflow on our fork: https://github.com/jmex-sw/IsaacTeleop/actions/runs/44007964864.
Could you approve the workflow runs for fa1b5e067?

@ivany-nv ivany-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks. The MOXIReceiverSDK_DIR check and the ungated core and tests are fixed in fa1b5e067. CI is green, and the 25 jmex tests now run and pass on x64 and arm64. LGTM.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants