kerfmaster pin list · Big capabilities
rulingA decision by Aristide or Jordan. True because it was decided; it can be superseded, but it cannot be stale.
Content last changed 2026-08-14 — computed from the item itself, not typed.
Contract — emit stage
This contract states a decision, not an implementation.
takes healing ?
makes dxf ?gcode
fails if A delivered coordinate carrying more than 4 decimals; a structural decision (merge / collapse / match) hinging on a difference below 0.0001 in.
Contract last changed 2026-08-14 — computed, not typed. Dated separately from the text above, so neither date can speak for the other.
The ruling (Jordan, 2026-08-14): “Remember when we said only 4 decimal places for gcode? I think the same thing can apply for healing. Millionths of an inch will never matter to us.” Prompted by the -74 circle-collapse fix, whose report was quoting scatters of 30 and 14 millionths.
What it means: the posted NC already speaks the shop's 4-decimal format (item 4a — 4 decimals max, measured off their own programs). The same resolution now applies to every healed COORDINATE we put on paper: emit.write_dxf_metacam (coords, radii, extents), dxf_r2000 (LINE/ARC/CIRCLE coords + radii), dxf_compat (the R12 shop writer). One count of the format is 0.0001" — a tenth of a thou; anything below that is noise and must never drive a structural decision (the -74 defect: an exact-match test let 0.00003" of arc-fit scatter decide circle-vs-two-arcs).
What it does NOT mean: internal math stays in doubles end-to-end — rounding happens at the emit boundary only, exactly as in the NC post. Angles (ARC 50/51, degrees) and bulges (dimensionless) keep their finer formats: they are not inches, and one count of a degree at a big radius is coarser than a count of the coordinate format. Tolerance GATES stay where their rulings put them (heal 0.010/0.001, circle collapse 0.001, close_tol 0.008) — this ruling sets the floor under them, not their values.
Verified on builds 075-080 (regenerated same commit): all six re-split, censuses unchanged (-74: 8 CIRCLE + outer POLYLINE; -75: 10 CIRCLE, 0 POLYLINE), every loop still closes exactly after rounding (shared vertices round identically — a 4-decimal grid cannot open a joint inside a polyline).