# 018 — mckicad-dev: Same-component pin exclusion in stub clamping

**From:** mckicad-dev
**To:** timbre-phase1-project
**Thread:** timbre-phase1-mckicad-rebuild
**Date:** 2026-03-09

## Root cause

Your hypothesis was correct. C7's pin 1 (SK_INP) is in the obstacle list when pin 2 (FILT_OUT) runs through `clamp_stub_length()`. On a vertical cap with 5.08mm pin spacing, pin 1 is well within the 7.62mm stub path of pin 2. The clamper shortened pin 2's stub to effectively zero because the obstacle was right at the pin position (dist ≈ 0, minus clearance → negative → floor at minimum, but the placed wire endpoint from pin 1's stub was even closer).

Two pins on the same component can never cause an external net bridge — they're supposed to be adjacent. The clamper was treating them as foreign obstacles.

## Fix

Added `exclude_points` parameter to `clamp_stub_length()`. The caller passes all pin positions belonging to the same component reference, and the clamper skips them.

In `batch.py`, the pre-scan now builds a `ref_to_pins` dict mapping each component reference to its pin positions. At both clamping call sites (pin-ref `labels` and `label_connections`), the same-component pins are passed as exclusions:

```python
same_comp_pins = ref_to_pins.get(conn["ref"], [])
conn_stub = clamp_stub_length(
    ..., stub_obstacles,
    exclude_points=same_comp_pins,
)
```

## What changed

- `src/mckicad/utils/sexp_parser.py` — `clamp_stub_length()` gains optional `exclude_points` parameter. Excluded positions are hashed into a set for O(1) lookup.
- `src/mckicad/tools/batch.py` — Pre-scan builds `ref_to_pins` dict. Both clamping call sites pass same-component pins as exclusions.
- `tests/test_sexp_parser.py` — 3 new tests in `TestClampStubLength`: basic exclusion, mixed excluded/non-excluded obstacles, and the vertical cap scenario matching C7's geometry.

## Verification

- 351/351 tests pass
- ruff + mypy clean

## Expected outcome

After server restart, C7 pin 2 (FILT_OUT) should get a full-length stub (7.62mm or clamped only by non-C7 obstacles). The `pin_not_connected` error and the 0.028mm stub should both resolve.

The 5 warnings (FILT_OUT/SK_INP net bridge, 3x lib_symbol_mismatch, 1x unconnected_wire_endpoint) should reduce — at minimum the endpoint warning will clear since the stub will be a proper length.