Add sandbox-verified fix bundle for customisable QidiBox firmware (9 fixes) — coded by Hermes AI, NOT hardware-tested

[kgrames]

Summary

Adds customisable_qidibox_firmware/sandbox-fixes-2026-08-05/ — a patch set on top of the community .so → .py swap with 9 fixes (6 fork-independent + 3 mainline-drift shims).

⚠️ IMPORTANT — testing status & authorship

• Coded by: Hermes, an AI coding assistant, with human review (2026-08-05).
• NOT tested on a physical QIDI Plus 4 printer. All fixes were verified in a Klipper software sandbox only (Klipper's CI file-output debug mode, virtual MCUs, no serial/hardware). The sandbox runs mainline Klipper; the real printer runs QIDI's fork (1.8.1), so fork-only API calls were guarded but not exercised on hardware.
• Use at your own risk. Keep your *.so backups, follow the install checklist, and test one change at a time. Community testing reports are very welcome.

Fix inventory

#	File	Fix	Class
1	aht20_f.py	4× None-guards on i2c reads (connect/CCP/AFE paths) — connect crash when a read returns None	fork-independent
2	box_extras.py	accept buffer_pin config option (auto-generated box.cfg always emits it; without this the .py swap fails to boot)	fork-independent
3	box_extras.py	E_UNLOAD: get_temp_by_num() None → 190 °C default (front-panel unload crash)	fork-independent
4	box_stepper.py	note_kinematic_activity() → get_last_move_time() (removed from mainline)	mainline-drift
5	filament_motion_sensor.py	QIDI-fork sensor + note_filament_present(eventtime, bool) signature	mainline-drift
6	box_extras.py	set_box_temp: box_max_temps KeyError + temp_info None guards (BOX_TEMP_SET crash with no box detected)	fork-independent
7	box_extras.py	TRY_MOVE_AGAIN: _00-suffix int-normalization + 2× None-temp guards (retry-button crash)	fork-independent
8	box_extras.py	heater.set_can_wait() guarded with hasattr (QIDI-fork-only API — TOOL_CHANGE_START crashes mainline)	mainline-drift
9	box_extras.py	filament_list_path config option (hardcoded /home/mks/... path broke non-QIDI hosts) + 3× None-guards (button_extruder_load ×2, QDE_004_009 retry path)	fork-independent
—	box.cfg macro	G1 E25 F300 → G1 E-25 F300 in UNLOAD_FILAMENT (stock unload extruded 25 mm instead of retracting — verified against QIDI's own generated template)	fork-independent

Install checklist (all items verified REQUIRED in sandbox)

1. box_macros.cfg must be present — registered CLEAR_FLUSH runs BOX_CLEAR_FLUSH; without it flushing silently no-ops (clogged nozzle / color bleed risk).
2. officiall_filas_list.cfg + filament_list_path in [box_extras] — temp lookups read this file; the hardcoded default path only exists on QIDI printers.
3. [filament_switch_sensor fila] section must exist (macros reference SENSOR=fila; missing → SET_FILAMENT_SENSOR error → shutdown).
4. [box_extras] pins (real, from QIDI's own template): b_button_pin: ^mcu_box1:PB1, b_endstop_pin: mcu_box1:PA9, e_endstop_pin: mcu_box1:PA10, buffer_pin: ^mcu_box1:PB0.
5. Box includes load AFTER heaters/extruder (box includes at the END of printer.cfg).

Full checklist + verification method: see the branch README.

What was NOT fixed (stock-firmware side)

• QDE_004_024 / M104 S0 temp-stability (issue I-12): emitted by the stock .so module, not the community .py. Sandbox confirmed the .py path handles M104 S0 + re-heat cleanly, but the .so behavior needs QIDI.
• Purge amounts (issue I-06): EXTRUSION_AND_FLUSH hardcodes ~180 mm of purge per change — separate config-level change.

How it was verified (reproducible)

• Klipper file-output debug mode (-d dict -o out -i gcode), synthesized STM32-pin dictionary, 3 virtual MCUs.
• 9 test files, all exit 0 with clean "Exiting" markers: boot, box command surface, unload macro (E-25 A/B), auto-reload + retry, tool change (TOOL_CHANGE_START/END + BOX_CHANGE_FILAMENT), purge (EXTRUSION_AND_FLUSH), temp stability (M104 S0 + re-heat), misc batch (RELOAD_ALL / CLEAR_RUNOUT_NUM / TIGHTEN_FILAMENT).
• Fork-only command warnings observed on mainline (fine on the real fork): QIDI_PROBE_PIN_2, BED_MESH_CLEAR, CLEAR_MOTION_DATA.

[kgrames]

Hi @vega-d (and any other maintainers) — gentle bump on this one, since the repo has been quiet for a few weeks.

Confirming it is ready for review from our side:

• All 9 fixes pass the full sandbox regression suite (boot, box config, unload, reload, toolchange, purge, temperature-stability, misc — every scenario exits 0 with clean shutdown). Fixes Add couple models, lid prop and exhaust adapter. #8 and Screw Sizes? #9 were added after the initial findings post and are covered by the same suite; full test matrix is in the PR.
• Class split still holds: Fixes improve chamber heating issue macro #1–Update README.md #3, Screws tilt adjust #6, Screws tilt adjust #7, Screw Sizes? #9 are fork-independent (relevant to QIDI stock 1.8.1 as well as mainline). Fixes Replace several old repo links to the new one #4, Screws tilt adjust #5, Add couple models, lid prop and exhaust adapter. #8 are mainline-drift shims only — QIDI's fork already has the correct behavior, so they are no-ops there, but they keep the community .py path from crashing on mainline-based configs.
• One install-checklist flag (F12): the registered CLEAR_FLUSH handler internally calls BOX_CLEAR_FLUSH, which lives in the community box_macros.cfg companion file — anyone swapping to the .py extras must keep that file present or flush silently no-ops. Documented in the PR body.
• Testing caveat unchanged: sandbox-verified only, not yet exercised on a physical Plus 4. Community hardware test reports are very welcome — happy to walk anyone through the test .gcode scenarios.

Happy to adjust code style, split the PR, add docs, or provide detail on any individual fix. Thanks for your time!