mirror of
https://github.com/maziggy/bambuddy.git
synced 2026-10-20 09:03:43 +02:00
fix(mqtt): close #1822 — promote H2S tray_now to 254 on all-external prints
H2S firmware reports tray_now=0 (the AMS's idle slot) throughout
external-spool prints instead of 254 like X1C/P1S/A1 do, so the
single-nozzle branch's 0-3 passthrough landed state.tray_now on slot
0 — UI highlighted AMS SLOT 1 instead of the external spool.
Usage credit was unaffected (#1276 covers that via ams_mapping).
The single-nozzle branch now checks _captured_ams_mapping (slicer-
captured per-filament mapping that the request-topic intercept
already tracks) before the existing P2S multi-AMS resolver. When
every entry is -1, the print uses ONLY the external spool, so
state.tray_now is promoted to 254.
Narrow on purpose: AMS-only [5] and mixed [5, -1] are NOT
overridden — we have no evidence H2S misreports mid-print swaps, and
trusting the firmware preserves correctness for users with multi-
filament setups. No-mapping prints (printer-screen start) fall
through unchanged.
This commit is contained in:
@@ -5884,3 +5884,87 @@ class TestAmsFilamentBackupHoldTimer:
|
||||
assert mqtt_client.state.ams_filament_backup is True
|
||||
# Hold timer still armed — sub-second push didn't reset it.
|
||||
assert mqtt_client._xcam_hold_start["print_option_auto_switch_filament"] == before_hold
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 2c. Single-nozzle H2S — external-spool tray_now override (#1822)
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
|
||||
class TestTrayNowH2SExternalSpoolOverride:
|
||||
"""H2S firmware reports tray_now as the AMS's idle slot (typically 0)
|
||||
instead of 254 when the active feed is the external spool.
|
||||
|
||||
Bambuddy detects the all-external case via the slicer-captured
|
||||
ams_mapping (every entry == -1) and promotes tray_now to 254 so the
|
||||
UI active-tray highlight matches the real feed.
|
||||
|
||||
The override is intentionally narrow:
|
||||
* only fires when ams_mapping is captured AND every entry is -1
|
||||
* does not touch mixed prints ([5, -1]) or AMS-only prints ([5])
|
||||
* does not fire when no ams_mapping is captured (printer-screen start)
|
||||
"""
|
||||
|
||||
@pytest.fixture
|
||||
def mqtt_client(self):
|
||||
from backend.app.services.bambu_mqtt import BambuMQTTClient
|
||||
|
||||
return BambuMQTTClient(
|
||||
ip_address="192.168.1.100",
|
||||
serial_number="TEST_H2S",
|
||||
access_code="12345678",
|
||||
)
|
||||
|
||||
def test_all_external_mapping_promotes_tray_now_to_254(self, mqtt_client):
|
||||
"""Reporter's scenario: H2S, single nozzle, captured ams_mapping=[-1],
|
||||
firmware sends tray_now=0 -> Bambuddy promotes to 254."""
|
||||
mqtt_client._captured_ams_mapping = [-1]
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 254
|
||||
|
||||
def test_multi_external_mapping_also_promotes(self, mqtt_client):
|
||||
"""Multi-filament print, every filament mapped to external. Still
|
||||
all-external -> still promotes."""
|
||||
mqtt_client._captured_ams_mapping = [-1, -1, -1]
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 254
|
||||
|
||||
def test_ams_only_mapping_does_not_override(self, mqtt_client):
|
||||
"""ams_mapping=[5] (AMS slot 5 only) -> firmware value trusted as-is.
|
||||
Without the all-external guard, this would falsely override real
|
||||
AMS-slot prints."""
|
||||
mqtt_client._captured_ams_mapping = [5]
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 0
|
||||
|
||||
def test_mixed_mapping_does_not_override(self, mqtt_client):
|
||||
"""Mixed mapping (AMS slot 5 + external): we have no evidence the
|
||||
firmware misreports mid-print swaps, so leave tray_now alone."""
|
||||
mqtt_client._captured_ams_mapping = [5, -1]
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 0
|
||||
|
||||
def test_no_captured_mapping_does_not_override(self, mqtt_client):
|
||||
"""Prints started from the printer screen (or before Bambuddy
|
||||
connected) have no captured ams_mapping. Behaviour unchanged from
|
||||
pre-#1822 — we accept the wrong value rather than guess."""
|
||||
mqtt_client._captured_ams_mapping = None
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 0
|
||||
|
||||
def test_empty_captured_mapping_does_not_override(self, mqtt_client):
|
||||
"""Empty list (defensive — should not happen in practice but
|
||||
all([]) returns True). Treat as no signal, not as all-external."""
|
||||
mqtt_client._captured_ams_mapping = []
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 0
|
||||
|
||||
def test_unloaded_after_print_still_resolves_correctly(self, mqtt_client):
|
||||
"""When the firmware unloads (tray_now=255), the override is skipped
|
||||
because the branch only fires for tray_now in 0-3."""
|
||||
mqtt_client._captured_ams_mapping = [-1]
|
||||
mqtt_client._process_message(_ams_payload(0))
|
||||
assert mqtt_client.state.tray_now == 254
|
||||
|
||||
mqtt_client._process_message(_ams_payload(255))
|
||||
assert mqtt_client.state.tray_now == 255
|
||||
|
||||
Reference in New Issue
Block a user