maziggy 0b8cc823e7 fix(queue): print a plate whose filaments are all on the external spool (issue #3087) ...

The reporter's P1S heated up, sat at Heatbed preheating for ten and a half
    minutes, then paused with 07FF_8012, "Failed to get AMS mapping table".
    Resuming only reheated it. Prints that fed from the AMS were fine.

    The plate was one filament of a seven-filament MakerWorld project, mapped by
    hand to the external spool. slice_info.config numbers filaments across the
    whole project, so the mapping for that plate is [-1,-1,-1,-1,-1,-1,254]: six
    placeholders and the spool holder. The command builder decides whether a print
    needs the AMS by asking whether the mapping is entirely external, and six -1s
    answer no. So the print went out as use_ams=true carrying a flat mapping of
    nothing but -1 -- 254 is deliberately never sent raw, the firmware reads it as
    AMS tray 0 -- which is exactly the mapping table the firmware then could not
    find.

    The builder cannot fix this itself. Down there a -1 is either padding for a
    filament this plate does not print, which is BambuStudio's own convention, or a
    slot that never resolved to a tray, and sending the second one to the spool
    holder is what #2589 exists to prevent. They are the same byte.

    The scheduler knows. extract_filament_requirements drops every filament with
    used_g <= 0, so it names precisely the slots the plate prints. When all of those
    are an explicit 254/255, dispatch now sends use_ams=false and the print runs.
    When one of them resolved to nothing, the flag is left alone and the firmware
    rejects the print as it does today -- deliberately, because that is the case
    where guessing would print a filament in the wrong material without saying so.

    Single-nozzle only, mirroring the reconcile in the command builder: on a
    two-extruder printer use_ams selects which nozzle to feed rather than whether to
    use the AMS, so an H2D with a spool on each side must keep the flag it was
    given. Judged generously from the model name and from live telemetry -- a second
    nozzle reporting a diameter, an extruder map, or more than one external feed --
    because a wrong yes only preserves existing behaviour while a wrong no would
    reroute the print. H2S stays single-nozzle (#1386).

    Nothing in the command builder changed. Its own reconcile keeps the exact
    semantics #2589, #2595 and #797 gave it, and now usually agrees with a decision
    that was already made one layer up. Where the file has no parseable filament
    list the mapping is left exactly as before, the same evidence-only convention
    as #2771, and the parse itself sits behind a check for anything external at all
    so an AMS-only print never opens the file.

    Covered end to end at the dispatcher, including the reporter's seven-filament
    shape, a plate mixing the spool holder with an AMS tray, a consumed slot that
    never resolved, both external feeds on a dual-nozzle machine, and a 3MF with no
    filament list at all.

2026-09-20 13:35:25 +02:00

..

app

fix(queue): print a plate whose filaments are all on the external spool (issue #3087)

2026-09-20 13:35:25 +02:00

tests

fix(queue): print a plate whose filaments are all on the external spool (issue #3087)

2026-09-18 17:55:06 +02:00

__init__.py

…