[gareth-morgan-nv]

Description

Draft. Stacked on #1461 (base is gmorgan/SensingLiveKit, so this diff is only the capture-backend rearchitecture). The Argus backend itself will be a further branch (gmorgan/Argus) on top of this one. There is no Argus SDK, Argus camera or Jetson available to the author, so none of the Argus path exists yet or has been tested. This PR does the part that does not need that hardware.

• Capture boundary (core/camera_capture.hpp): a small vendor-neutral CameraCapture interface, a move-only FrameLease, and a tagged pixel type (borrowed CUDA RGBA for SIPL, leased NV12 DMA-BUF reserved for Argus). The release and DMA-BUF contracts are written down in the header.
• SIPL behind the interface (SiplCapture): a thin adapter with no extra copy; the output still copies the borrowed frame out after its pacing check, as before. SensingCamera and LiveKitVideoOutput take the interface's frame type; other pixel types are rejected until an encoder input exists. Behaviour is intended to be unchanged.
• Backend selection (--camera-backend=auto|sipl|argus): chosen once at startup from each backend's probe result. An explicit request must match or fail with that backend's own reason (no silent fallback); auto takes the only match and otherwise names the override. Only SIPL is built in, so argus fails with "this build does not include that backend".
• Encoder: a stopped capture thread is now reported by poll() instead of looking like a healthy encoder; removed an unused per-frame stamp.
• Unit tests for the selection rules and the frame lease (Catch2, no CUDA/SIPL needed).

Not done (needs an Orin or an Argus camera)

• The Argus capture module and the encoder's NV12 DMA-BUF input.
• A supported per-frame correlation between encoder input and output on the Jetson. Units are paired by submission order today, so an encoder-internal drop can shift frame ids; this needs target-header verification before it is fixed.
• Distinguishing "no input buffer yet" from a terminal driver error on the encoder's output dequeue (depends on the driver's errno contract).
• Optional dynamically loaded backend modules.

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

• Selection and frame-lease tests: 12 cases pass when built standalone with g++ against a Catch2 checkout. The CMake test target itself was not configured.
• New and edited C++ files pass clang-format-14; the SIPL adapter, selection and main.cpp pass a g++ syntax check. jetson_encoder.cpp could not be syntax-checked here (needs the Multimedia API headers).
• Nothing has been built or run on a Jetson since these changes.

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

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: NVIDIA/IsaacCapture
• Branch: gmorgan/SensingLiveKit
• File: .coderabbit.yaml
• Commit: 33b09ed
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: bf5009cd-ae0f-4a70-a7eb-ddfd830eb797

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

[github-actions]

✅ Preview deployed: https://NVIDIA.github.io/IsaacCapture/preview/pr-1428/

Built from commit 840e767. It may take a minute for the updated preview to appear.

[gareth-morgan-nv]

Closing as superseded: the camera path is now the camera_streamer architecture (SIPL/Argus sender, AVS1 over TCP, PC receiver and LiveKit bridge; see #1473-#1479). Not merged; the branch and commits are kept.

🤖 Generated with Claude Code

[gareth-morgan-nv]

This PR is superseded by the camera_streamer PRs:

• camera_streamer: add SIPL camera support to the sender #1473 camera_streamer: add SIPL camera support to the sender
• camera_streamer: map SIPL sensor timestamps to sender CLOCK_MONOTONIC and add an AVS1 timing record #1474 camera_streamer: map SIPL sensor timestamps to sender CLOCK_MONOTONIC and add an AVS1 timing record
• camera_streamer: add a UDP clock exchange responder to the sender #1475 camera_streamer: add a UDP clock exchange responder to the sender
• examples/camera_streamer: add a camera-free mock sender and publish the first decoded picture #1476 examples/camera_streamer: add a camera-free mock sender and publish the first decoded picture
• camera_streamer: add cross_build.sh and deploy_sender.py development tools #1477 camera_streamer: add cross_build.sh and deploy_sender.py development tools
• examples/camera_livekit_bridge: publish receiver frames to LiveKit #1478 examples/camera_livekit_bridge: publish receiver frames to LiveKit
• examples/camera_streamer: carry a qualified sensor capture time to LiveKit #1479 examples/camera_streamer: carry a qualified sensor capture time to LiveKit

🤖 Generated with Claude Code