[hougantc-nvda]

Description

Review stack · 2/5: #8787 → #8788 → #8789 → #8790 → #8791

Review only this layer. These PRs target main; the comparison excludes predecessors.

Connect live Isaac Capture tracking and recorded MCAP input to kitless simulation through #8787. SO101 gains CPU OvPhysX and Newton selection, plus the current full-home, clutch and gripper integration. Its IK task resets to nominal joints without reset noise on Kit and kitless; the separate joint-control task retains randomized resets.

• Package-owned entrypoints provide live teleoperation and MCAP replay; existing scripts delegate to them. Kit keeps --xr, and kitless uses --ovxr.
• A declared optional IsaacTeleopCfg.prepare_environment callback finalizes task settings before simulation creation. Dependent Enable G1 and GR1T2 tasks with kitless XR #8789 supplies the humanoid implementation; generic launchers do not infer optional backend hooks.
• Replay honors recorded START/RESET by default. --start_on_replay explicitly supports recordings without START; reset does not rewind the recording or become terminal STOP. HDF5 workflows remain on Kit.
• Cleanup failures remain visible, and unretired borrowers keep their resources. Breaking behavior and migration are recorded in the changelog.

Narrow shared fixes cover empty CPU Warp arrays, nonblocking graph capture in the Newton backend, and simulation-context retention after failed asynchronous drain. Kit retains headless XR updates without changing ordinary headless on-demand rendering.

The runtime guide links canonical workstation, client, network and license prerequisites. In-process control/configuration tests are separate from subprocess launch smoke tests.

Installation: use published isaaccapture==1.7.93a1 from NVIDIA’s Python index with the scoped OVXR installation. The guide creates the optional ovxr selection; project-wide Kit/provider defaults are unchanged.

Type of change

• New feature
• Bug fix
• Documentation update

Testing

The combined stack passed 877 tests across 33 affected test files, with no failures or skips. This includes the native Newton graph-capture/replay/reset tests, renderer/lease contracts, recorded-control gating, camera cleanup failures, registered task selection and CPU collision checks, plus the inherited upstream noise-reset cases. Formatting, all five changelog boundaries and the warning-as-error documentation build pass. These are cumulative integration results; individual PR boundaries received static/API and changelog checks. Runtime and performance scope is detailed in #8791.

Remaining qualification: native Kit execution and MCAP-file playback were not rerun; control/replay owner tests substitute external I/O. Native consumer-destruction failure is not established by normal shutdown. Cold renderer attachment, intermittent GR1T2 first-Play recovery and SO101 cube retention remain open. Existing USD, render-product GPU-assignment and native runtime warnings are retained; this is not a warning-free qualification.

Checklist

• Contribution guidelines read
• Changed-file formatting/lint and Python 3.12 changelog checks pass
• Public documentation and release fragments updated
• Affected behavior and cumulative IWER regression checks complete within the scope above
• Native Kit and physical-headset qualification at this revision
• No new warnings

[hougantc-nvda]

@greptile Please review the latest changes in this PR.

[greptile-apps]

Confidence Score: 4/5

[Medium impact] The optional OVXR route connects teleoperation, simulation, and an external render worker.

Update the publication-failure test double before merging; both parameter cases otherwise fail at the new material lookup.

Findings

1.  Publication tests fail early ▶

Summary

Adds kitless live teleoperation and MCAP replay, SO101 support, and ordered XR cleanup.

Since the last review:

• Live tracking and recorded input can drive scenes shown through OVXR.
• OVRTX resources can be leased to an external renderer owner.
• The SO101 absolute-IK task supports CPU OVPhysX and a full gripper home pose.
• CPU arrays, Newton graph capture, and simulation cleanup support the kitless runtime.
• Headless Kit applications keep updating the scene while XR is active.

Previous unnumbered threads were rechecked without posting duplicates. The callback uses validate_live(), and Kit updates both anchors from the filtered pose. hougantc-nvda disputed the replay Reset finding because supported Capture forced requests only defer during shutdown; no retry change was added.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[Task changes material] --> B[Mark channel dirty]
    B --> C[Check whether publication is needed]
    C --> D[Reserve scene ordinal]
    D --> E[Write materials and scene]
    E --> F[Wait for material reads]
    F --> G[Seal scene ordinal]
    G --> H[XR worker renders]

Loading

_{Reviews (4) · Last reviewed commit: "Merge updated prerequisite for #8788" · Reviewed by Greptile}

[hougantc-nvda]

@greptile Please review the current head (62a1200) again. Recheck the fixes and remaining issues against the published code, including the actual interfaces and callers at this PR boundary.

[hougantc-nvda]

@greptile please review this PR again at the latest head (01fc06561c0d65c5dbe75fe56e5e18b7172dcd91), including the changes since your previous review.

[isaaclab-review-bot]

Isaac Lab Review Bot

Needs attention: Design and architecture (2), Scope and over-engineering (5), Compatibility and deprecation (4), Implementation (5), Style consistency (5), Test quality (8). Other reviewed categories look good.

• PR description: The body cites non-head validation and itself as dependency documentation; replace those references, relocate stray template prose, and provide reproducible Newton capture performance evidence or mark that hot-path risk unresolved.
• Design and architecture: Backend preparation uses string-built hooks, while hand presence reaches through private device lifecycle state; declare task preparation explicitly and expose hand presence through the public device API.
• Scope and over-engineering: Minimal implementation: keep render_lease.py for irreducible ownership lifecycle; merge publication_buffers.py into ovrtx_renderer.py; delete unused packing_table.py and action_policy.py, moving required policy into isaac_teleop_cfg.py/entrypoints; keep ov_libraries_xr.py and ovxr_session.py for Capture adaptation/bootstrap. Suggested PR breakdown: (1) shared runtime fixes; (2) renderer lease/publication; (3) live and replay OVXR entrypoints; (4) SO101 behavior and docs, each with focused owner-boundary validation.
• API: API changes look good.
• Compatibility and deprecation: Breaking changes: ordinary teleop imports optional isaaclab_ov, cleanup errors propagate, pipelined resets become edge-triggered, and registered SO101 initialization changes; isolate dependencies or provide complete transition bridges.
• Implementation: Replay injects a local START before consuming recorded controls, advancing pre-START frames and defeating recorded pauses; remove the injection and let MCAP control events exclusively drive activity.
• Style consistency: New teleop modules use absolute intra-package imports contrary to repository style; restore relative imports and address the associated typing, header, stale prose, and private-access violations.
• Test quality: Added tests repeatedly execute mutated AST fragments, grep source, self-compare exports, restate configuration, and duplicate weaker helper boundaries; replace them with imported public or owner-boundary behavior tests.

Needs rework. Posted 29 inline findings.

Automated review; human maintainers own approval decisions.

[isaaclab-review-bot]

Isaac Lab Review Bot

Needs attention: Design and architecture (2), Scope and over-engineering (5), Compatibility and deprecation (4), Implementation (5), Style consistency (5), Test quality (8). Other reviewed categories look good.

• PR description: The body cites non-head validation and itself as dependency documentation; replace those references, relocate stray template prose, and provide reproducible Newton capture performance evidence or mark that hot-path risk unresolved.
• Design and architecture: Backend preparation uses string-built hooks, while hand presence reaches through private device lifecycle state; declare task preparation explicitly and expose hand presence through the public device API.
• Scope and over-engineering: Minimal implementation: keep render_lease.py for irreducible ownership lifecycle; merge publication_buffers.py into ovrtx_renderer.py; delete unused packing_table.py and action_policy.py, moving required policy into isaac_teleop_cfg.py/entrypoints; keep ov_libraries_xr.py and ovxr_session.py for Capture adaptation/bootstrap. Suggested PR breakdown: (1) shared runtime fixes; (2) renderer lease/publication; (3) live and replay OVXR entrypoints; (4) SO101 behavior and docs, each with focused owner-boundary validation.
• API: API changes look good.
• Compatibility and deprecation: Breaking changes: ordinary teleop imports optional isaaclab_ov, cleanup errors propagate, pipelined resets become edge-triggered, and registered SO101 initialization changes; isolate dependencies or provide complete transition bridges.
• Implementation: Replay injects a local START before consuming recorded controls, advancing pre-START frames and defeating recorded pauses; remove the injection and let MCAP control events exclusively drive activity.
• Style consistency: New teleop modules use absolute intra-package imports contrary to repository style; restore relative imports and address the associated typing, header, stale prose, and private-access violations.
• Test quality: Added tests repeatedly execute mutated AST fragments, grep source, self-compare exports, restate configuration, and duplicate weaker helper boundaries; replace them with imported public or owner-boundary behavior tests.

Needs rework. Posted 29 inline findings.

Automated review; human maintainers own approval decisions.

[hougantc-nvda]

@isaaclab-review-bot review

[isaaclab-review-bot]

Isaac Lab Review Bot

Needs attention: Design and architecture (3), Scope and over-engineering (3), Compatibility and deprecation (5), Implementation (1), Style consistency (7), Test quality (4). Other reviewed categories look good.

• PR description: Update the body: this diff adds the OVXR adapter, #8788 is self-referential, and performance results use another revision; correct ownership and links, then benchmark the capture-stream change at head or mark it unmeasured.
• Design and architecture: cleanup_context=False retains SimulationContext while launchers still close, leaving live borrowers without their runtime; retain launcher ownership until drain completes and replace backend-name dispatch with typed configuration.
• Scope and over-engineering: Minimal implementation: keep render_lease.py, ov_libraries_xr.py, and ovxr_session.py for irreducible ownership lifecycles; merge publication_buffers.py into ovrtx_renderer.py; delete action_policy.py and its unused hooks; centralize render-admission retry in OvLibrariesXrSession. Suggested PR breakdown: (1) shared Warp/Newton/launcher/Kit fixes; (2) OVRTX lease and CPU publication; (3) live/replay OVXR, SO101, and documentation.
• API: API changes look good.
• Compatibility and deprecation: Breaking changes: replay startup, SO101 reset/calibration, run_teleoperation positional binding, cleanup error propagation, and exported partition visibility changed without deprecation bridges.
• Implementation: prepare_teleop_backend imports optional isaaclab_ov on every teleop launch, breaking Kit-only installations; defer that import until the resolved physics type requires it.
• Style consistency: The README pipeline factory reuses one pipeline instance instead of constructing fresh session state; document a real factory and separate retargeter factory.
• Test quality: The smoke-test module now contains roughly 700 lines of unit tests, obscuring CI ownership; move them to focused unit modules and remove non-discriminating or duplicated reset/export cases.

Needs rework. Posted 23 inline findings.

Automated review; human maintainers own approval decisions.