[Binyaminp]

Overview: Adds a new "Edge Enhancement" visual preset, tuned for stronger edge preservation, available on all D400 devices.

Tracked on [RSDSO-21893]

Why

D400 devices lacked a preset tuned for stronger edge preservation in the depth map.

Summary

• Added RS2_RS400_VISUAL_PRESET_EDGE_ENHANCEMENT to rs2_rs400_visual_preset (appended before _COUNT, no existing values changed)
• Added edge_enhancement() in presets.cpp/.h, populated from a validated EdgeMap tuning JSON for D400 devices
• Wired the new preset into apply_preset() in advanced_mode.cpp and into the rs2_rs400_visual_preset_to_string() mapping
• get_preset_max_value() in d400-device.cpp now returns EDGE_ENHANCEMENT for all D400 devices (not just D435/D435I) so the preset is reachable via RS2_OPTION_VISUAL_PRESET everywhere
• Added ds_advanced_mode_base::is_preset_supported(), checked against both firmware-registered feature support and device pid, so a preset a device doesn't support (e.g. REMOVE_IR_PATTERN on anything but D400/410/415/460) is no longer listed as a selectable option in the Visual Preset dropdown, instead of only failing at apply time
• option_model::get_combo_labels() now skips option values with no description instead of assuming every value in [min, max] is valid

Test plan

• Built realsense2 core library (Release, x64) — compiles cleanly, no new warnings
• Built and launched realsense-viewer with the change
• Manually verified "Edge Enhancement" appears in the Visual Preset dropdown in Advanced Mode on a D435i and applies correctly; Depth Clamp Max reads 65536 as expected
• Verified on real hardware (D405, a passive/laser-less D400 SKU, FW 5.16.0.1): REMOVE_IR_PATTERN is correctly hidden from the option's enumerated values, and applying EDGE_ENHANCEMENT (laser_state=0) succeeds with no exception

---

🤖 Generated by AI

[sysrsbuild-gh-agentic]

Auto-generated review by rs-agentic bot. Found 2 bugs and 1 note below.

[sysrsbuild-gh-agentic]

🤖 [rs-agentic] Reviewed new commit — one new bug noted inline; prior PID-coverage fix approved.

[Nir-Az]

get_preset_max_value() now returns EDGE_ENHANCEMENT for every D400, so REMOVE_IR_PATTERN (5) is inside [min,max] on devices that were previously capped below it. VALIDATE_RANGE in rs2_set_option (src/rs.cpp:949) no longer rejects it — rejection now depends on apply_preset() being reached.

The viewer half is handled by skipping null labels in get_combo_labels, but option_model::is_enum() (common/option-model.cpp:247) still returns false when any value in range lacks a description, so is_enum() gates the combo upstream and Visual Preset renders as a slider instead of a dropdown on D435/D455/D405. tools/dds/dds-adapter/lrs-device-controller.cpp:178 bails the same way for DDS.

Suggested:

1. is_enum() — skip undescribed values rather than rejecting the option. Checked every enum option in the tree; all describe their full range, so nothing else changes rendering. Same one-line change at lrs-device-controller.cpp:178.
2. Add the is_preset_supported() check to advanced_mode_preset_option::set() (src/ds/advanced_mode/advanced_mode.cpp:1120), right after is_valid(value), so an unsupported preset fails deterministically instead of depending on apply_preset() — today an advanced-mode-disabled device throws wrong_api_call_sequence_exception first.

(1) is the blocking one; (2) is hardening.

🤖 Generated by AI

[Binyaminp]

✅ Fixed in d3a1621, both points.

1. is_enum() now returns true if at least one value in range describes itself, instead of bailing to false on the first null — same one-line fix applied to get_option_enum_values() in lrs-device-controller.cpp:178 (return {} → continue).
2. Added the is_preset_supported() check to advanced_mode_preset_option::set() right after is_valid(value), before the advanced-mode check.

Verified on the D405: attempting RS2_RS400_VISUAL_PRESET_REMOVE_IR_PATTERN now throws immediately with "set(advanced_mode_preset_option) failed! Device does not support preset Remove Ir Pattern" (the new deterministic error), and EDGE_ENHANCEMENT still applies cleanly with no regression.

🤖 Generated by AI

[sysrsbuild-gh-agentic]

🤖 [rs-agentic] Reviewed the latest synchronize commit. Two new bugs found in the changes.

[sysrsbuild-gh-agentic]

🤖 [rs-agentic] BUG: The default: throw was removed from the inner switch(device_pid) inside the REMOVE_IR_PATTERN case. is_preset_supported() and the dispatch switch now maintain two separate PID lists — if a PID is added to is_preset_supported() but omitted from the dispatch switch, the check passes, the switch matches no case, and apply_preset() silently continues with an uninitialised/default preset instead of throwing. Add back default: throw (or a default: assert(false)) to make the mismatch visible at runtime.

[sysrsbuild-gh-agentic]

🤖 [rs-agentic] BUG: If current_value equals a value whose label is nullptr (hidden for this device), the continue skips the current_index = labels.size() assignment. The caller receives a stale/wrong current_index, misidentifying the currently-active option in the combo box. Consider checking i == current_value before the continue and either logging a warning or falling back to index 0 when the live value has no visible description.

[Nir-Az]

LGTM. Enum appended before _COUNT so no ABI change, and apply_preset()/get_value_description() share one supported-check instead of duplicating the pid logic.

Checked the relaxed is_enum() against every enum option in the tree — all describe their full range, so nothing else changes between slider and dropdown rendering.

One thing still worth eyeballing before merge: the viewer itself on a device without Remove IR Pattern, that Visual Preset still draws as a dropdown. Verification so far is API-level on a D405.