[aceRage]

URGENT safety fix. On 2.4.6.2, a QIDI printer with a QIDI Box was started on a plate sliced as PLA (PLA and ASA in the Box) with no Box mapping sent, and printed the ASA at PLA settings.

Root cause

The QIDI Box firmware feeds T<n> from the slot named by its saved variable value_t<n>. Only the sending slicer sets that variable. #611 sent SAVE_VARIABLE enable_box / value_t<n> from exactly one place: PrintHostSendDialog::qidi_start_script() (PrintHostDialogs.cpp), which is empty unless m_device_status.qidi is true. The printer then reuses whatever its previous print left in value_t<n>. These paths start a print with no mapping:

1. No dialog at all: hub / phone app / Stream tab send and archive reprint. RemoteSend::prepare_host (RemoteSend.cpp ~644-720) and prepare_from_record (printhost branch, ~1287-1351) set post_action = StartPrint, and run_host (~1582) calls host->upload() directly. That upload is Octo/Moonraker print=true. There is no probe and no Box script, and nothing on the desktop pops up. This matches "the pre-send dialog never popped up".
2. Desktop dialog, probe failed (address down, HTTP 404, timeout, or a non-Moonraker reply). refresh_device_status() gets qidi = false. The dialog is then the plain one, with no Box rows and no QIDI options, and "Upload and Print" is enabled. The owner reproduced this on an X-Max 4 preset.
3. Desktop dialog, no device row. build_device_ui() returns early when m_devices.empty() (PrintHostDialogs.cpp ~553). That happens with a preset print_host and the device deleted from the list (migrate never re-imports it). Nothing is probed.
4. Desktop dialog, a filament left "As sliced". No value_t<n> is sent for it, so it still loads from the stale slot.
5. "Print all": plate_filaments_for_send only marked the current plate's filaments as used.

The sidebar Print / Send buttons, print plate and print all all go through send_gcode_legacy, so they always show the dialog. There is no "don't show again" or remembered post-action for print hosts. QIDI profiles have no host_type, so they default to octoprint (the OctoPrint class talking to Moonraker), which is probed.

What the owner's logs and screenshot show (PLA calibration fed from the ASA slot)

• The dialog probe timed out at 3 s on a real QIDI on the LAN ("Status: unknown ... Operation timed out after 3000 milliseconds"). With that timeout there are no Box rows and no options.
• In 2026-10-19-18-38-46.log.0:
  • At 20:51 the QIDI options script timed out after 60 s and the send stopped, which is correct.
  • At 20:54 the options script succeeded, so m_device_status.qidi was true, and the print started. Only the bed-leveling call failed ("Klippy Write Request Error"). The "sent ... (Box mapping X)" line was info level, so the mapping that went out is not in the log. It is now logged at warning level.
• The calibration did not use a separate path. The Calibration menu (Plater::calib_pa / calib_flowrate / calib_temp, Plater.cpp ~20160 / 20664 / 20720) only builds a normal project with new_project + add_model. Its object prints on filament 1 (T0, value_t0) and is sent with the same sidebar Print -> send_gcode_legacy -> dialog.
• QIDIStudio 2.07 never reads multi_color_controller (the string is not in its binary). It reads the Box from save_variables (box_count, filament_slot<n>, color_slot<n>) plus the printer's config/officiall_filas_list.cfg. The value quoting VALUE="'slot%s'" and the 0-based box*4+slot numbering match ours.
• The most likely mechanism: the Box is detected (and on), but no slot named a readable material (slots_have_filaments false), so no rows are shown. A filament can also be left "As sliced" because its type did not match. In both cases qidi_start_script sends SAVE_VARIABLE enable_box VALUE=1 alone, Moonraker answers OK, and the printer keeps value_t0 = slot<ASA> from the earlier ASA print.

Fix

• PrintHostDevices::qidi_send_policy(QidiSendFacts) is one pure function that every path uses. Its results are NotApplicable / Start / UploadOnlyBusy / UploadOnlyNoProbe / NeedsMapping / ConfirmTypeMismatch. The helpers are qidi_box_capable_model (profile: Q2/Q2C, X-Max 4, X-Plus 4, X-Plus 5), filament_types_match and qidi_auto_mapping (material-first, as the dialog's auto-match).
• Desktop dialog:
  • A Box-capable preset with no device list gets its own address as the device (as Elegoo Link already does).
  • "Upload and Print" asks the policy:
    • Unreadable Box (probe failed, or the printer answered without multi_color_controller): Upload only (the default and the auto-answer), Retry, or an explicit "Print anyway".
    • Unmapped used filament: the print is refused, with the list.
    • Slot holds another material (PLA vs ASA): a warning with "Change the slots" / "Print anyway".
  • The note under the printer says "Couldn't read the QIDI Box: the printer would use its last saved slot mapping". The QIDI options are not offered blind, because they need the printer to answer and the queue's idle check would stop the send anyway.
  • The mapping is read through mapping_choice_positions(), so it also works with the Bambu-style swatch map for third-party send dialogs (CANVAS, U1, QIDI Box) #625/Happy Hare and AFC slot reading for the Moonraker send dialog #612 swatch panel.
• Hub / phone / Stream tab / archive reprint (arm_qidi_guard + qidi_guard_start):
  • Applies to any Moonraker-speaking print-host target, or a Box-capable model.
  • Probes the printer and auto-maps by material.
  • If the policy says Start, it sends the Box script (enable_box + value_t<n>) before the upload. Otherwise it refuses with the reason and nothing is sent.
• "Print all" counts every plate's filaments.
• Reading the Box:
  • The probe waits 10 s for a Box-capable model (3 s for everything else).
  • When multi_color_controller names no materials, the probe reads the Box the QIDIStudio way (save_variables + officiall_filas_list.cfg).
  • enable_box / box_count in save_variables also mark QIDI firmware with a Box.
• Box default: an empty or unreadable Box defaults to on. Printing from the toolhead spool (enable_box 0) feeds whatever is in the toolhead, so it is only ever an explicit untick.
• Read-back: after the Box script, the desktop queue (PrintHost.cpp) and the hub send read save_variables back. If enable_box and every value_t<n> do not match what was sent, they refuse to upload.
• Material mismatch confirmation for every host whose slots name materials (the CANVAS dialog included). A PLA filament going to an ASA slot asks "Change the slots" / "Print anyway".

Other hosts: unchanged. The exceptions are one extra read-only probe on a remote print to an octoprint-type host, and "print all" marking the filaments of every plate.

Same hole elsewhere (reported, not fixed)

• Elegoo CANVAS remote send: prepare_host sends no extended_info, so no slot_map. The printer gets the start without a mapping, exactly as before the CANVAS work. This is a different firmware: there is no saved stale variable, but the result is still unmapped. It needs a CANVAS read in run_host.

• U1: no hole. The remote path is the LAN snapmaker kind, with auto_match plus start_print_mapped, and the desktop uses the connect page.

• Device tab file list "Print" (RemoteFileGridModel::StartPrint -> MoonrakerFiles::start_print) on a QIDI: it starts a stored file on the stale mapping, the same as the printer's own touchscreen. Follow-up.

• Flashforge IFS dialog: it has its own slot dialog and does not use these rows, so it is not covered by the mismatch confirmation.

Unverified on hardware (nobody here has a QIDI Box)

• The officiall_filas_list.cfg layout ([fila<id>] type =, [colordict] <id> = #RRGGBB) is inferred from the property-tree keys in the QIDIStudio binary.
• So is the assumption that a slot with filament_slot<n> > 0 is loaded.
• If either guess is wrong, the slots read as unknown and the print is refused (NeedsMapping). A wrong slot is never started, because the read-back confirms what the printer saved.

Compatibility with #625 / #612

The guard reads mapping_choice_positions() and m_mapping_filament, which the swatch panel answers. Expect textual conflicts in PrintHostDialogs.cpp: update_mapping_note (#625 F14 removed its early return), rebuild_mapping_rows, and the QIDI note in refresh_device_status. #616's async probe would need confirm_qidi_start to run only after the probe finishes. #620 touches init() (the Upload and Print handler).

Tests

tests/slic3rutils/qidi_send_policy_tests.cpp covers the policy matrix, including "profile says QIDI Box, probe failed" and the owner's 404 repro. It also covers the reported PLA/ASA plate, PETG not in the Box, busy, Box off or absent, and the U1 not being affected.

Hand test (owner, no QIDI): with a QIDI X-Max 4 preset and a dead address, check that the note shows, that Upload is the default, and that Upload and Print shows Upload only / Retry / Print anyway.