[aceRage]

Description

Depends on #625 (merge #625 first) via #619. Merge the Creality stack (#618 → #616 → #619) then this.

#625 adoption: direct-rack (K3) printers opt in with third_party_map_read_only(). mapping[i] is warning_slot_position() (identity T-number) so tiles, highlight and mapping_warning() agree (F13). filament_mapping() stays empty; start stays opGcodeFile. K3 mapping_warning() is kept next to #625's third_party_mismatch_warning(). T1–T3 counts only lines whose previous significant line is ; custom gcode end (allowing blanks, ;_FORCE_RESUME_FAN_SPEED, and M106/M107 including P-indexed); a preceding T0 block cannot credit a later T1–T3 (F10). Slot.name tile labels apply only when direct_rack (F1).

Stacked PR. Base is cursor/creality-cfs-sync-d523 (head of the Creality stack: #618 → #616 → #619). Must land after #619. Do not merge onto main first.

Draft Creality K3 (model code F039) support. The K3 is on pre-order (not shipping). It is a Klipper 4-nozzle KliTek toolchanger: four independent nozzles, each fed from its own filament holder, one printing at a time, ~5 s swaps, no CFS.

This fork already handles that machine class for the Snapmaker U1 (single_extruder_multi_material=false, four nozzle_diameter entries, filament i → physical Ti). The K3 reuses that path. Creality Print's filament_map / filament_is_mixed machinery is not ported. No shared slicer code is changed in a way that would alter U1 G-code.

Screenshots/Recordings/Graphs

None. No hardware. The send dialog shows K3 holders read-only (what's loaded in each nozzle) and states that T0–T3 are as sliced.

Art (F9): Creality K3_cover.png and Creality K3_thumbnail.png are K2 Plus placeholders. They need real K3 art before this is more than a draft.

Tests

Catch2 cases were added in existing files only (no CMake edits). The full [CrealityK3] slice case was not compiled — this environment has no build/, /opt/edgeslicer-deps, or /opt/edgeslicer-app. The T-count helper was compiled standalone and passed.

• tests/slic3rutils/printhost_devices_tests.cpp: F039 is a known direct-rack model (not CFS, not K1-family); resolve_model("", "Creality K3") → F039; gcode root is the stack default /mnt/UDISK; opGcodeFile start; parse_direct_rack_response accepts four 1-slot racks or one 4-slot box and skips the type-1 external rack; status_from_direct_rack is never CFS; holder labels are Nozzle N (T{n}); mismatch helper warns on empty / type+colour (tolerant RGB distance) / filament 5+ with no nozzle. Identity fill: warning_slot_position for read-only mapping.
• tests/libslic3r/test_wipe_tower_no_sparse_gcode.cpp [CrealityK3]: four-cube K3 system-profile slice asserts the ported start sequence (START_PRINT EXTRUDER_TEMP=, BED_TEMP=, M104 S140, M104 S220 T, G1 X120 Y0 E15), END_PRINT / G1 X130 Y255, T0–T3 tool lines, and a K3 change-filament marker in the ; custom gcode end adjacency immediately before each T1/T2/T3. Only T lines whose most recent significant line is ; custom gcode end (allowing blanks, ;_FORCE_RESUME_FAN_SPEED, and M106/M107 including M106 P2 S255) are counted. Start G-code T and M104 T3 are not counted. A preceding T0 block cannot credit a later T1–T3 (F10). Marker: G1 Y… Z… F30000 (the template's Y±5 dodge) or I0.86 J0.86 P1 (spiral fallback). Not a bare Z…F30000 (travel z-hop uses that too).

orca_extra_profile_check.py --vendor Creality --check-filaments --check-materials --check-obsolete-keys --check-assets exits 0. Warning count is 714 on this branch and on cursor/creality-cfs-sync-d523. No printed line names K3.

---

Review 4/3 (da5e12a325) — panel-adoption B1 / F1

• B1. count_k3_change_t_lines also skips M106/M107 (including P-indexed, e.g. M106 P2 S255) between ; custom gcode end and T, because CoolingBuffer replaces ;_FORCE_RESUME_FAN_SPEED with those commands. Synthetic cases: marker then M106 S128 then T2; marker then M106 P2 S255 + M107 then T1. Start T3, M104 T3, and T0 no-credit remain. Standalone helper compile passed.
• F1. ds.label = s.name only when direct_rack, so it cannot override Happy Hare and AFC slot reading for the Moonraker send dialog #612's 1A–1D grouping for QIDI/HH/AFC. Comment on use_third_party_map_panel(): once Happy Hare and AFC slot reading for the Moonraker send dialog #612 merges it must be cfs || direct_rack || qidi_box || happy_hare || afc.

Review 2/3 (dbefe65b9a)

• B1 / F6 redo. Tool-change G-code is written before T. The test now looks in the nearest preceding custom-gcode block for the K3 dodge G1 Y… Z… F30000 or I0.86 J0.86 P1, and requires one such block per T1/T2/T3.

Review 1/3 follow-ups (8e4d32d9d5 on 35d4c1caf9)

• F6. Superseded by the 2/3 redo above (the after-T scan was wrong).
• F7. Holder colour compare uses colour_rgb_dist_sq (the same RGB helper auto_match_index uses). Near-identical shades (about 32 RGB units) do not warn. Filament 5+ emits “has no matching nozzle” instead of being ignored.
• F8. Send-dialog note and mismatch lines are _L() / format_wxstr literals so xgettext can extract them. Status.note stays English for Catch2.
• F9. Cover and thumbnail are K2 Plus placeholders and need real K3 art (listed above).

Review 1/3 blockers (35d4c1caf9)

• B1. Direct-rack UI is honest. Holders are read-only. No dropdown, no closest-colour prefill. filament_mapping() returns empty so history / Snapmaker unload cannot record a fake remap. Start stays opGcodeFile. Note says mapping is set when slicing (assign filaments to extruders); T0–T3 go to nozzles 1–4; send does not remap G-code. Type/colour mismatch vs the T-number holder is a warning only.
• B2. Added Creality K3_cover.png and Creality K3_thumbnail.png (K2 Plus stand-ins; see F9). Dropped obsolete keys: filament_load_time / filament_unload_time from Hyper PLA; overhang_speed_classic / overhang_totally_speed from the K3 process files.
• B3. Creality.json version 02.03.02.06 → 02.03.02.07.
• F1. Bed model/texture are K2/Hi 260×260: creality_hi_buildplate_model.stl + creality_k2_buildplate_texture.svg (not the K1 235×250 assets).
• F3. Holder labels are Nozzle 1 (T0) … Nozzle 4 (T3).
• F5. [CrealityK3] asserts the K3 start/tool-change sequence (see Review 2/3 for the tool-change half).

Validator output

$ python3 scripts/orca_extra_profile_check.py --vendor Creality --check-filaments --check-materials --check-obsolete-keys --check-assets
[INFO] Checking profiles ...
… pre-existing obsolete filament_load_time / filament_unload_time on K1/K2/SPARKX/Ender profiles (no K3 files) …
==================== SUMMARY ====================
[INFO] Checked vendors     : 1
[SUCCESS] Files with errors   : 0
[WARNING] Files with warnings : 714
=================================================
exit 0

Same command on cursor/creality-cfs-sync-d523 (#619): errors 0, warnings 714, same printed obsolete-key set. No new warnings.

---

Sources (Creality Print, AGPL)

Inspected on CrealityOfficial/CrealityPrint (current default branch at the time of this work):

Item	Where
Machine + model	resources/profiles/Creality/machine/Creality K3.json, Creality K3_model.json (nozzle_diameter 0.4×4, single_extruder_multi_material=false)
Process	resources/profiles/Creality/process/0.20mm Standard @Creality K3.json (and siblings)
Filament	resources/profiles/Creality/filament/Hyper PLA @Creality K3.json (and Generic PLA/PETG/ABS/ASA/TPU)
Scope / send revert	doc/creality-k3-current-direct-rack-scope-plan.md; SendToPrinter revert c709a2aaf (K3-specific filament_maps / open_cfs start was reverted; start is opGcodeFile)
Device panel	PrinterBoxFilamentPanel.cpp (model == "F039")
M109 / K3 gate	MachineVender::is_creality_k3 used only for OozePrevention::post_toolchange M109
F039 notes	doc/bugfixes/F039-K3-*.md
Vendor index	Creality Print Creality.json K3 entries

Upstream OrcaSlicer has no K3 profile.

What was adapted

• Profiles under resources/profiles/Creality/ (machine model + 0.4 nozzle, four processes, six filaments) and surgical inserts in Creality.json after the K2 SE counterparts. Schema adapted to EdgeSlicer: scalar nozzle_type / nozzle_volume; host_type / printer_agent = crealityprint; printer_model = Creality K3 (the stack preset fallback). Start / end / change_filament_gcode ported from Creality Print and credited in the machine JSON. tool_change_temprature_wait=0 (profile-gated M109 fix → M104). purge_in_prime_tower=0 on the machine. machine_tool_change_time=5.
• Model detection (CrealityPrintModel.hpp): direct_rack_models {F039 → K3} separate from cfs_capable_models. is_known_model includes F039 only (not the display name). resolve_model reverse-maps Creality K3 / K3 → F039. omit_upload_path("F039") is false. G-code root: stack default /mnt/UDISK (see unverified).
• Send / device: no new WebSocket start message. K3 stays on opGcodeFile + enableSelfTest. #618 deadline/cancel rules unchanged. Probe uses /info + boxsInfo; F039 is parsed by parse_direct_rack_response (does not reuse CFS 4-slot padding). If holders are reported they are shown read-only via the shared Bambu-style swatch map for third-party send dialogs (CANVAS, U1, QIDI Box) #625 panel; filaments stay T0–T3 as sliced. No "Use CFS". Calibrate before printing is kept because Creality Print puts enableSelfTest on generic opGcodeFile.
• Slicer: no shared wipe-tower / OozePrevention C++ changes (Creality Print's WipeTowerCreality zhop is not ported). U1 / Snapmaker files untouched.

Unverified (needs a real K3)

• G-code root. Creality Print's F039 sources do not name a K3-specific path. Used the stack default /mnt/UDISK. Flag this on first hardware test.
• Holder / boxsInfo schema. Parser accepts both "four type-0 boxes, one material each" and "one type-0 box, four materials" and skips type 1. On-wire layout is unknown.
• Network protocol. Assumed the same port-9999 opGcodeFile as other non-CFS Creality machines after the SendToPrinter revert. No K3 start-message capture.
• No hardware. No printer, no send, no print.
• Catch2 slice test not compiled. T-count helper compiled standalone; [CrealityK3] slice binary not built.
• Creality Print's K3 wipe-tower C++ (zhop) was not ported; only profile settings (enable_prime_tower, purge_in_prime_tower=0, ooze / standby).
• Cover/thumbnail are K2 Plus placeholders and need real K3 art.

Real-K3 smoke checklist

When a K3 is available:

1. Load Creality K3 0.4 nozzle + 0.20mm Standard @Creality K3 0.4 nozzle + four filaments. Confirm Check profiles / profile validator is clean.
2. Slice a 4-colour model. Confirm T0–T3, START_PRINT EXTRUDER_TEMP= / M104 S140 / M104 S220 T, a G1 Y… Z… F30000 (or I0.86 J0.86 P1) custom block immediately before each T1–T3, prime tower, and M104 (not wait-M109) on tool changes.
3. Confirm a Snapmaker U1 4-colour slice is byte-identical to cursor/creality-cfs-sync-d523 (this PR must not change U1 G-code).
4. Connect the printer. Confirm /info model is F039 and the preset fallback Creality K3 → F039.
5. Confirm the send dialog: no "Use CFS"; "Calibrate before printing" present; holders read-only via the shared map panel if reported; note says as-sliced T0–T3; no mapping dropdown; a type/colour mismatch vs the T-number holder warns only; filament 5+ warns it has no nozzle.
6. Confirm gcode lands under /mnt/UDISK/printer_data/gcodes/ (or record the real root).
7. Send a single-colour then a 4-colour job. Confirm start is opGcodeFile (no colorMatch / multiColorPrint / CFS fields) and the printer starts.
8. Confirm a ~5 s nozzle swap and that T0–T3 as sliced print the right colours (mapping is set in the slicer).
9. Cancel a send mid-probe / mid-start and confirm #618 deadline/cancel still applies (no hang).
10. Run libslic3r_tests "[CrealityK3]" and printhost_devices_tests (F039 / direct-rack sections) after a real compile.
11. Replace the K2 Plus cover/thumbnail placeholders with real K3 art.

  

[aceRage]

Code Evaluator review: re-review 3/3 @dbefe65b9a, FAIL (1 blocker, test only). The 3 re-review limit has been reached, so this goes to the upstream reviewer with the open findings below. The code itself is clean: no libslic3r, U1 or CFS changes, and the profile validator exits 0 with no new warnings.

B1, a one-line test fix: the [CrealityK3] count check (test_wipe_tower_no_sparse_gcode.cpp :777-778) also counts the start G-code's bare T3. With a prime tower on a non-Bambu printer and priming at 0, GCode.cpp:3435-3438 uses the last used extruder (3 with four extruders, ToolOrdering.cpp:1403-1406) as initial_no_support_extruder (:3442-3443). The K3 start G-code emits T[initial_no_support_extruder] ('Creality K3 0.4 nozzle.json':96), which is a bare T3 outside any custom block. So k3_blocks_before_t comes out one short of tool_changes and :778 fails even when every change runs. To fix it, ignore T lines before the first ; custom gcode start or the first layer marker, or only count T lines whose most recent non-blank line is ; custom gcode end (allowing ;_FORCE_RESUME_FAN_SPEED in between).

Open follow-ups:

• F10: a T1-3 change with no block of its own, straight after a T0 change, could wrongly get credit for the T0 block.
• F9: Creality K3_cover.png and _thumbnail.png are K2 Plus placeholders and need real K3 art.

Still owed: Edge CI, MSVC Release+Debug, Catch2 run, a real K3 send and print (send protocol, /mnt/UDISK root and boxsInfo holder layout are unverified), and a U1 byte-identity check. It must land after the Creality stack (#618 -> #616 -> #619).