[AviaAv]

Two unrelated fixes found while investigating RSDEV-12697 (which turned out not to be a bug - the difference there is a viewer-side colorizer setting, not recorded data).

1. Playback dropped the recorded sensor options (ros2_reader.cpp)

read_sensor_options() scans option ids in ascending order, so a value re-written mid-recording is out of order and starts a second, partial scan - whose result then overwrote the full snapshot. Now the first snapshot is kept; mid-recording changes still arrive during playback as serialized_option messages.

Measured on a D435I: 25 options recorded, 1 restored. Depth Units was among the lost ones, so playback fell back to a hardcoded 1mm - 10x off on a 0.1mm device (verified by setting a D455 to 0.0001).

2. The viewer discarded a saved "Histogram Equalization Enabled" (subdevice-model.cpp)

subdevice_model re-applies Visual Preset after restoring the config, and every preset also assigns equalization - so a saved 0 was replaced at startup and could not be turned off persistently. The saved value is now re-applied afterwards.

Test

pytest-record-and-stream.py now toggles the emitter mid-record and asserts no recorded option goes missing on playback. Fails on development and passes with the fix, on both D435I and D585.

[sysrsbuild-gh-agentic]

🤖 Auto-generated by rs-agentic bot

Reviewed new commit f2e0add — the expanded re-apply loop in subdevice-model.cpp is clean: supports() guard, range check, and try/catch are all in place, and equalization is applied last (after min/max) matching the observer ordering. No issues found.

[AviaAv]

Merging as last commit is just an update to the test + previously approved + LibCI run passed with all changes