Developer Test Command¶

_MMU_TEST is Happy Hare's grab-bag developer command - one gcode command, registered like any other, with roughly two dozen unrelated sub-tests selected by which parameter you pass. Its own source comment calls the tests "raw," and that's the right way to think about it: there's no extra safety net beyond what's described below, and several sub-tests exist specifically to provoke bugs rather than avoid them.

Why it's hidden, not disabled¶

The leading underscore is Klipper's plain convention for "don't list this in help" - it doesn't show up in MMU_HELP, and Happy Hare doesn't document its parameters anywhere in the normal command flow. That's the only thing hiding it: there's no separate developer-mode build flag, no Kconfig option that has to be enabled first. _MMU_TEST is registered and fully callable on every Happy Hare install, same as MMU_LOAD or MMU_STATUS. The only real gate is the same one every command has - MMU ENABLE=1 if the unit is currently disabled.

Unlike every real command, _MMU_TEST is deliberately excluded from the user-facing Command Reference - not something a regular user should stumble onto while looking up a real command. Its generated parameter list instead lives on Developer Command Reference, alongside the other internal/developer-only commands. This page is the why and when, grouped by how much you should trust running a given option without thinking twice.

Introspection - safe to run anytime¶

These just report state; none of them move anything.

_MMU_TEST GET_POS=1

Filament pos state: 4 (LOADED_ENCODER)

_MMU_TEST GET_POSITION=1

Filament position: 683.2

_MMU_TEST DUMP_ACTIVE_SENSORS=1

active_sensors={'mmu_gate_0': True, 'mmu_gate_1': False, ...}

_MMU_TEST DUMP_MCU_ENDSTOPS=1

mmu_gate_0(mmu:mmu,PA3,140234...)               Steppers: gear0

SENSOR=1 dumps what the sensor manager believes is fitted before/after a given filament position, without actually moving anything - useful for working out why a load/unload logic branch didn't trigger the sensor check you expected:

_MMU_TEST SENSOR=1 POS=4 GATE=0 LOADING=1

check_all_sensors_before(4,0)=True
sensors before=['mmu_gate_0']
check_all_sensors_after(4,0)=False
sensors after=[]

Two more worth knowing:

UPDATE_STATUS={dict} overrides what get_status() reports to Mainsail/Fluidd/KlipperScreen with a hand-supplied dict, without touching any real state - handy for testing a UI panel's rendering of an edge-case status without physically reproducing it. UPDATE_STATUS=OFF removes the override.
NFC_READ=1 simulates a tag scan with no reader hardware attached at all, injecting at the exact point a real reader hands off - every reader-level guard is bypassed by construction, but the real downstream feature gates (nfc_deep_read, spoolman_support) still apply. Useful for exercising the NFC → Spoolman path on a bench with no tags or readers wired up: _MMU_TEST NFC_READ=1 UID=E2003412 GATE=0 DEEP=1 MATERIAL=PETG.

Moves real hardware - bench-test it, don't run it mid-print¶

These genuinely turn motors, so treat them like any other motion command: know where the selector and filament actually are before you run one.

WRAP_CURRENT=1 exercises the gear/extruder current-wrapping context manager (MOTOR= PERCENT=) - nested calls, to confirm currents restore correctly when one wrap ends inside another.
SEL_MOVE=1 / SEL_HOMING_MOVE=1 / SEL_LOAD_TEST=1 drive the selector directly (MOVE=, SPEED=, ACCEL=, LOOP=), bypassing the higher-level gate-selection logic - for isolating a selector-only problem from everything built on top of it.
SYNC_LOAD_TEST=1 and REALISTIC_SYNC_TEST=1 both hammer gear/extruder sync and movement in a loop with randomized parameters (LOOP=, ENDSTOP=, SELECT=, SERVO=/WAIT=) - the closest thing to a stress test in this command. Both need a hot extruder and either a toolhead sensor or an explicit ENDSTOP=; read the warning each one logs on startup before running it. REALISTIC_SYNC_TEST additionally cross-checks tracked position against reported position every loop and raises MmuError the moment they disagree.
SET_POS=, SET_POSITION=, SET_ACTION=, FILAMENT_POS=, FILAMENT_DIR= force internal state directly - filament position state, raw filament position, the current action, and so on - without the movement or checks that would normally produce that state. Good for reaching a specific state to test a downstream code path; bad for anything that then assumes the physical filament matches what you just told Happy Hare.
ADJUST_ENCODER=/SET_ENCODER= nudge or overwrite the encoder's tracked distance directly, independent of any real filament movement.

Provokes known bugs on purpose - core-team debugging only¶

These exist to reproduce specific, already-known failure classes in Happy Hare's stepper-sync/homing logic. They're not something a regular contributor needs, and a couple have sharp edges:

TTC_TEST=1 / TTC_TEST2=1 / TTC_TEST3=1 each drive a different randomized homing/move pattern (LOOP=, MIX=, DEBUG=, WAIT=) specifically to try to provoke a Klipper "Too Then Close" (TTC) timing violation in the MMU/extruder stepper sync path.
STEPCOMPRESS_TEST=1 does the same for step-compression errors (LOOP=, MOTOR=, STOP_ON_ENDSTOP=), wrapping each move in DebugStepperMovement for extra diagnostics on failure.
QUIESCE_TEST=1 runs a fixed, hand-written sequence of sync/unsync transitions and raw extruder moves that has previously exposed "quiescing" bugs around sync-state changes. Needs a hot extruder and a toolhead sensor (or dummies).
SYNC_STATE= (compression/tension/both/neutral) fabricates sync-feedback sensor events without touching real hardware, using temporary phony sensors if the real ones aren't fitted or enabled. SYNC_STATE=loop is refused outright - its result-gathering is a busy-wait that blocks the single reactor greenlet, so the events it's waiting for can never actually arrive. That's a real, permanent limitation of this test, not a bug to work around by retrying.

Sequence timing, without a real print¶

RUN_SEQUENCE=1 and RUN_CHANGE_SEQUENCE=1 run through the pre/post-unload and pre/post-load macro hooks (timing each phase via the same statistics machinery a real toolchange uses) without an actual print running. RUN_CHANGE_SEQUENCE's NEXT_POS=next also exercises the toolhead-parking/return-position logic. Both support ERROR=/PAUSE= to force a mid-sequence failure and confirm the pause/statistics path behaves. Useful for timing or debugging your own macro hooks (see Macro Customization) without slicing and printing a real file each time.

Feeding fake telemetry to autotune¶

NOTE_LOAD_TELEMETRY=1/NOTE_UNLOAD_TELEMETRY=1 call the same calibrator methods a real load/unload calls at the end of its move, but with hand-supplied LENGTH=/TRAVEL=/RATIO= numbers instead of a real measurement - the only way to exercise autotune_rotation_distance's correction logic repeatedly without physically loading and unloading filament every time. TRAVEL defaults to LENGTH (a bare call is inert by design); name a different TRAVEL to provoke a specific correction.

The rest¶

A handful of narrower one-offs, each self-contained:

Parameter	What it does
`SYNC=`	Force a specific gear/extruder sync mode directly (`0-3` or a name)
`RUNOUT=`	Enable/disable runout arming for a gate, bypassing the normal flow
`CALC_PURGE=1`	Run the purge-volume-by-color calculator against a few fixed test cases
`SEND_PRINTING_EVENT=`	Fire the `mmu:printing`/`mmu:not_printing` event by hand
`ACTIVATE_FLOWGUARD=`	Call FlowGuard's activation/deactivation hooks directly
`DUMP_UNICODE=1`	Print every special glyph the console/log UI can render, for a font/terminal check