janikandClaude Fable 5
6d8634802d
Build PCM package / build (push) Successful in 7s
tests / ubuntu-latest · py3.11 (push) Successful in 53s
tests / ubuntu-latest · py3.13 (push) Successful in 1m43s
tests / archlinux:latest (push) Successful in 39s
tests / debian:12 (push) Successful in 49s
tests / fedora:latest (push) Successful in 1m1s
tests / ubuntu:24.04 (push) Successful in 1m26s
tests / NixOS (FHS wrapper from docs/NIXOS.md) (push) Has been skipped
New EXPERIMENTAL dialog option (default off): after the solve, copper whose |J| is below a threshold (default 10% of the mean |J| over all solved copper cells - mean, not max, since contact-corner spikes would dwarf a max-relative threshold) is vectorized into filled graphic polygons on TRIM_LAYERS (User.5..User.8, configurable), one polygon per region so Edit > Convert can turn one into a rule area by hand. Areas are printed and the polygons saved to low_current_copper.json. The mask -> polygon step is the 0.5 contour of the binary field via contourpy (already in every venv as matplotlib dependency), padded so regions touching the raster edge close, simplified with Douglas-Peucker at 0.4 cells: staircase bevels collapse, one-cell-wide strips survive. Specks under TRIM_MIN_AREA_MM2 are dropped. Explicitly a suggestion, not a safe cut list (docstring, dialog and README all say so): copper carries little current BECAUSE the rest carries it, so removal redistributes |J| - the constant-density optimizer that iterates this to convergence is future work. board_io: the create/delete-with-status-surfaced helpers are now generic (_create_items_checked / _remove_items_checked) and shared between reference-image overlays and trim polygons. Tested on the dev stack (3.13) and the Python 3.9 mac-stack venv, 148 passed each; contourpy 1.3.x has identical API on both. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>