ROSBag: parse the RealSense image footer instead of guessing it

[TianHengZhuang]

Problem

RealSense SDK bags append four extra bytes to every image message: a little-endian float32 holding the depth scale (0.001 for 16UC1 depth). The reader currently strips four bytes from every connection as soon as one realsense_msgs connection is present, so the value is discarded instead of being exposed, and recordings whose messages are self-contained (for example Intel's published single_depth_color_640x480.bag) lose the last four bytes of every image.

Change

• sensor_msgs Image and CompressedImage are self-describing, so the payload end is computed from the field layout and only the bytes past that boundary are treated as a footer.
• The footer is unpacked as a little-endian float and attached to the yielded Image as depth_units.
• Trim amounts are probed in the order 0, 4, 8 with the untrimmed buffer first, so self-contained messages are never truncated. rosbags SerdeError is retried the same way as AssertionError.
• Connections that are not images fall back to the untrimmed buffer when a trailer was inferred from another connection.

CDR encoded messages (ROS 2 wire format inside a ROS 1 container) use the same approach with CDR alignment.

Tests

tests/create_ros_fixture.py now also writes test_ros1_realsense.bag (four byte footer on every image, 0.001 on the depth stream) and test_ros1_realsense_notrailer.bag (same topics, no footer).

TestRosBagRealSense covers topic detection, decoding of both streams, depth_units == 0.001 on depth, 0.0 on colour, and the absence of the attribute for self-contained recordings.

The rosbag tests need the ros extra and the generated fixtures. I checked the decode path of both new bags and of the existing ROS 1 and ROS 2 fixtures with rosbags directly; the full suite runs in CI.

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 28 complexity · 2 duplication

Metric	Results
Complexity	28
Duplication	2

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}