cohdl (conol-ai) errors

4 recurring problems engineers report in cohdl (conol-ai), ranked by impact.

#122 by errors trackedECAD

Recurring cohdl (conol-ai) errors, by impact

  1. 1A footprint library's JLC frame rows rest on readings taken by eye from screenshots, and the DFN1006 diode polarity is not confirmed. measurements.json has no dots for D1 to D4 or C9 to C12. A reviewer asked for each row's evidence to say so, and now it does in a note field no generator emits, while the rotation and polarity data a fab depends on still comes from a person squinting at a preview.ECAD · Agent failure · 1 report66
  2. 2A part library holds six diode parts in draft because their LCSC numbers are not in its sourcing map, and fetching them needs approval. Polarity was checked by eye on a render with a 180 degree rotation applied, and the SS34's 3D model is labelled SMAJ12A, so the check covers polarity only, not identity.ECAD · Agent failure · 1 report66
  3. 3ESP32 module antenna areas were emitted only as silkscreen outlines across twenty generated footprints, so PCB tools could route or place parts inside them, and IPC-2581 export now refuses these footprints because it cannot carry the restriction.ECAD · Agent failure · 1 report66
  4. 4Supplier part pages carry only package text, not a footprint id. C7972 and C7722 both read SOT-23-5, yet need rotations of 180 and 270 degrees, so all 9,629 mapped chip resistors and MLCCs would each need their own placement preview, and 3,442 with no stock can never get one.ECAD · Agent failure · 1 report66

How it fails

  1. Agent failure4

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