You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Placement check reports 'inside' for repeated instances that are off the plate #93
The placement check reports the wrong result when one object is used by more than one build item (a repeated instance).
backend/snapstudio_core/placement.py builds its table of placements with dict(_BUILD_ITEM.findall(root)), which keeps only the last transform for each object id. plate_placement._printable_only then gives every instance of that object the footprint of the last one.
A project with one cube object and two build items that use the same object id, one at X=500 and one at X=100 (the first is far outside the U1 bed). With a model_settings.config part record present, plate_placement.assess() returns:
off_plate count 0
both items at position (105, 105)
the summary "Every object sits inside the Snapmaker U1's printable area."
The instance at X=500 is not inside the printable area.
Impact
A false "inside" result for repeated instances. It is a claim that a printable area check passed when it did not.
Notes
Found during plan review for the read-only 3D project inspection. That work does not fix this; it disables engine-derived out-of-bounds highlighting for objects with more than one build item and says so.
Fix direction: key placement by build item (ordinal and component path), not by object id; add a repeated-instance fixture and compare per-instance results with the scene's world bounds.
Orca's own "add instance" output was not checked, so how common this layout is in real projects is not known.
Verified after the merge of #102 (849e156). On merged main (tree identical to the verified #102 head) test_repeated_instance_placement.py and test_repeated_component.py pass, including the issue's reproduction in both orders (X=500 and X=100: the X=500 instance is reported off the plate, the other inside), both-off, all-inside, per-component, nested-transform, per-plate and scene-world-bounds cases (50 tests across the placement files). Full backend suite on the merged tree: 3649 passed, 49 skipped; all 8 CI checks green. Coverage limit by design: projects whose physical size the file does not establish (non-millimetre unit, rotated or scaled item) report the placement check as not checked instead of a per-instance verdict. Orca's own 'add instance' output was still not checked.
Activity
fix(placement): key by part, never treat unjudged instances as fine (#93
fix(bed-fit): unknown stays unknown, every object checked, measure on…
fix(placement): bound the footprint read; number instances per (part,…
fix(bed-fit): real printable area, every failing object, fresh cache (#…
fix(placement): printable area not travel, no OK for unjudged or tall…
fix(bed-fit): identified printer's printable area, locked caches (#91 …
fix(placement): content-keyed Ready Now cache, height from printing p…
fix(bed-fit): unidentified printer is unknown, placed height checked (#…
fix(placement): below-plate parts, torn Ready Now reads, named printer (
fix(bed-fit): size from printing parts only, sunk parts flagged (#91 …
fix(placement): printing roles propagate down through assemblies (#93 …
fix(bed-fit): sizes use the shared role rule (#91 repair 6)
fix(placement): role-aware overall extents; Orca precise-seam helper …
fix(bed-fit): role-aware extents and precise-seam roles across all co…
test(placement): the cross-consumer test also drives service.bed_fit …
Merge pull request #102 from DVOpenLabs/fix/93-repeated-instances
DVOpenLabs commentedon Oct 22, 2026
Verified after the merge of #102 (849e156). On merged main (tree identical to the verified #102 head) test_repeated_instance_placement.py and test_repeated_component.py pass, including the issue's reproduction in both orders (X=500 and X=100: the X=500 instance is reported off the plate, the other inside), both-off, all-inside, per-component, nested-transform, per-plate and scene-world-bounds cases (50 tests across the placement files). Full backend suite on the merged tree: 3649 passed, 49 skipped; all 8 CI checks green. Coverage limit by design: projects whose physical size the file does not establish (non-millimetre unit, rotated or scaled item) report the placement check as not checked instead of a per-instance verdict. Orca's own 'add instance' output was still not checked.