mirror of
https://github.com/maziggy/bambuddy.git
synced 2026-09-30 11:12:35 +02:00
maziggy
eee94ce9c8
The K value on an AMS slot card went blank after a while and came back after a
backend restart. It is not the MQTT merge: that preserves a tray's k correctly.
H2-series trays have no k to preserve. Verified against the H2 wire capture in
logs/vp_wire -- every tray reports cali_idx and nothing else -- so the number on
the card is resolved from that index against the printer's calibration table in
state.kprofiles, and that table was a single global list.
An extrusion_cali_get response is the complete table for one nozzle diameter,
and the printer answers whoever asks; BambuStudio's queries arrive on the same
report topic we subscribe to. Every response was assigned straight to
state.kprofiles, so any one answer stood for the whole printer. The nightly
GitHub backup asks for 0.2, 0.4, 0.6 and 0.8 in turn and finishes on 0.8, which
holds nothing on a 0.4+0.6 machine: logs/bambuddy.log records exactly that at
17:15 on 2026-08-25, and the table was empty from then until something refilled
it. Responses are now bucketed by the diameter they describe, so an empty answer
for a size the printer does not have clears only that size. The three assign
paths that look an index up by nozzle_diameter get the same fix for free -- they
were quietly finding nothing whenever the last response was for another nozzle,
which is what spoolman_inventory has been logging as a stale kp.
Bucketing makes state.kprofiles a union, and cali_idx is numbered per nozzle, so
the index alone no longer identifies a profile. The REST serializer has keyed on
(extruder, cali_idx) since c5e005586; the WebSocket one still keyed on the index
alone, which meant the first render of a card could be right and every update
after it wrong. Both now share one resolver. It goes through the extruder the
slot feeds, and where that does not single out one profile -- a single-nozzle
printer that has been swapped, so both its tables sit under extruder 0 -- it
falls back to which diameters are actually fitted. Where neither settles it the
card shows nothing, because a blank space is a smaller error than confidently
printing the other nozzle's number. Deliberately no loosening to a bare cali_idx
lookup on a miss: that is the cross-nozzle bleed the extruder keying was added
to stop.
Nothing read the table on connect, which is the other half of the report. It
arrived by luck -- a visit to Profiles or Configure Slot, a backup, or the
printer answering someone else -- so a Bambuddy nobody had opened showed a card
with no K values at all, and "restart and they come back" was the printer
happening to broadcast rather than anything we did. It is now read once per
connection, on the same latch the stale-print reconcile uses. Only the fitted
diameters are asked for, one request on a single-nozzle printer and two on a
dual; probing the four sizes blind is what the backup does and what blanked the
table. The edge is gated on a nozzle diameter being known as well as on the
state being known, because the first push_status is what makes the state known
and does not always carry the nozzle fields -- latching there would spend the
connection's one attempt on a printer that could not yet say what was fitted.
Adopting an unsolicited table now logs at debug. It was the quietest way for the
card to change underneath us and there was no way to see it in a support bundle.
Three test files gained nozzles=[] on their PrinterState stubs. The field has
always been on the dataclass; the connect edge is simply the first thing on that
path to read it.