copperline (iamdarshg) errors

6 recurring problems engineers report in copperline (iamdarshg), ranked by impact.

#71 by errors trackedECADMCAD

Recurring copperline (iamdarshg) errors, by impact

  1. 1Eight curated boards that were legal going in came out of the router with copper DRC violations (56 on one, 11 on two others). Routes from each epoch get committed with no local legality check or rollback; the final verifier refuses COMPLETE only after the bad copper is already in.ECAD · Agent failure · 1 report74
  2. 2New tools/verify_rules.py is billed as per-board manufacturability verification but fails open: if router analyze fails or returns no nets, the width loop skips every trace and prints 0 below required width instead of not checked; it checks file existence, not return codes, and never clears last run's report JSONs, so a failed route can be scored against an old board's report, and main() always exits 0 even with illegal DRC or failed RF/thermal validators.ECAD · Agent failure · 1 report74
  3. 3On a 160x100 mm, 12-layer, 842-pad board that started at 0 DRC errors, the BGA escape stage committed 170 vias and stubs and produced 90 copper clearance violations, with only 405 of 842 pads connected, and each stub checked when planned, never against copper already committed.ECAD · Agent failure · 1 report74
  4. 4A PCB routing benchmark dedupes boards by rounded size, layer count, net count and pad count, before checking legality. Distinct legal boards get dropped, and an illegal board can knock out a later legal one with the same counts, biasing completion and legal-rate claims.ECAD · Agent failure · 1 report66
  5. 5KiCad 10 writes a pad's net as (net 'GND') with no number, and copperline parses the first atom as a number, so it fails with 'bad number '+12V' in pad net'. The reporter sees it on 7 upgraded boards and on boards from another KiCad 10 generator. KiCad's own Specctra export also writes a boundary form the DSN reader does not take.ECAD · Agent failure · 1 report66
  6. 6The synthetic routing corpus generator puts different-net pads on top of each other (0.28 mm and 0.32 mm overlaps) before routing starts. The harness records input_legal but still counts those boards, so invalid inputs show up as router failures in the metrics.MCAD · Agent failure · 1 report66

How it fails

  1. Agent failure6

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.