[petercorke]

Summary

The Intel RealSense SDK follows every image message in a bag with 4 extra bytes (a float32 depth scale). The reader handled that by assuming 4 trailing bytes on every message of any bag with a realsense_msgs connection, and only when release="auto". Tested on the real bags in a development checkout, this had three problems:

Bag	Old reader	New reader
RealSense SDK recording (4-byte footers)	default ROSBag(path) reads 0 frames (SerdeError: ROS1 message size mismatch); auto works but discards the footer	reads all frames; depth_units == 0.001 on the depth stream
Intel's sample bag single_depth_color_640x480.bag (realsense_msgs connections, no footers)	auto returns 0 frames, silently (it removes 4 real bytes from every image)	reads all frames
the 12 other bags (ROS 1 and a ROS 2 MCAP recording)	fine	byte-identical output

What changes

• sensor_msgs/Image and CompressedImage (ROS 1 and CDR, either byte order) are parsed to the end of their data array, checking length == height * step. Only those bytes are deserialized; anything after them is the footer. Pixels are unchanged.
• What to make of the footer is decided by context, because the bytes cannot decide it (about half of all 4-byte patterns are a positive finite float32). In a bag recognised as RealSense, a positive finite float becomes the image's depth_units attribute (metres per unit) and four zero bytes (a color image) mean no scale. Anything else after an image (the same 4 bytes in an unrecognised bag, another length, a negative or non-finite value) is ignored with a RuntimeWarning, once per topic. CDR padding is silent.
• Every other message type must be exactly the length its type says, and is no longer retried with 0/4/8 bytes removed. All 68,889 non-image messages in the 13 real bags decode at their exact length, so that guess was not needed.
• SerdeError is caught as well as AssertionError (current rosbags raises SerdeError).
• The reasoning is in the docstrings of ROSBag, _image_payload_end, _footer_depth_units and _image_depth_units.

Related issue

Refs #151. The depth_units= / invalid_is_nan= conversion options are not in this PR; they will follow separately.

Credit

The approach of finding the end of the data from the message layout follows #154 (@TianHengZhuang), and the technique for building a synthetic RealSense-style bag in the tests follows #155 (@theluckystrike). Both are credited as co-authors on the commit. Thank you both.

Verification

• New tests/test_ros_realsense.py (332 tests): the layout parser, footer decoder, synthetic RealSense-style bags written to a temporary directory by tests/create_ros_fixture.py (ROS 1 / CDR LE / CDR BE, with and without footers and realsense_msgs), unexpected trailing bytes, exact-length decoding of other messages, and checks against real bags that skip when bags/ is absent. No binary fixtures are committed.
• Run against the old reader, 7 of 12 synthetic cases come out wrong or incomplete.
• pytest tests: 1367 passed, 44 skipped.
• Old vs new reader on every bag in bags/ (14 files): identical pixels on every frame.

Checklist

• PR title follows Conventional Commits
• Tests pass locally (pytest)
• Added tests for this change
• Docstrings updated

🤖 Generated with Claude Code

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 77 complexity · 0 duplication

Metric	Results
Complexity	77
Duplication	0

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.}