kicad-mcp-server

Author SHA1 Message Date

Author	SHA1	Message	Date
Eugene Mikhantyev	496be317e9	test: close mirror-fixture gap exposed by post-fix audit The eeschema-ground-truth fixture (_build_mirror_case) passed 'mirror': 'x' to ComponentManager.add_component, which silently drops the kwarg — so the resulting .kicad_sch had no (mirror x\|y) token, eeschema rendered an unmirrored symbol, and our pin coords (also unmirrored) tautologically matched. The mirror tests were GREEN both before and after the rotation/mirror fix in `7e67cb9`, providing zero regression coverage for the mirror semantics. Fix: - _build_mirror_case now applies the mirror via the same low-level helper (WireDragger.update_symbol_rotation_mirror) that rotate_schematic_component uses, with a guard assertion that the written file actually contains (mirror x\|y). - Two new pin-down unit tests in test_add_schematic_component.py document and lock down ComponentManager.add_component's silent-drop behavior for mirror, so the next person to touch that path knows to update the eeschema-truth fixture if they grow real mirror support. Verified: with the production fix at `7e67cb9` reverted, the kicad-cli mirror tests now go RED (previously they stayed GREEN regardless). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	2026-05-03 22:40:34 +01:00
Eugene Mikhantyev	f7660e15ad	test: encode eeschema kicad-cli ground truth (currently RED) Update existing rotation/mirror assertions and add a kicad-cli-grounded regression suite. eeschema's pin transform is verified against: - rotation: TRANSFORM(0,1,-1,0) for rot=90 — CCW in screen Y-down - mirror x: SYM_MIRROR_X = TRANSFORM(1,0,0,-1) — negates internal Y - mirror y: SYM_MIRROR_Y = TRANSFORM(-1,0,0,1) — negates internal X Updated assertions: - test_pin_locator_y_flip: rotated cap pin 1 X 153.81 → 146.19 - test_move_with_wire_preservation::test_resistor_rotated_90: pin 1 X 103.81 → 96.19 - test_hierarchical_pad_net_map::test_90_degree_rotation: swapped UP_NET / DN_NET label coords New file tests/test_pin_world_xy_eeschema_truth.py: - 4 parametrized diode tests via kicad-cli netlist (oracle = eeschema) - 2 parametrized resistor mirror tests via kicad-cli netlist - 3 pure-math pin_world_xy assertions for rot=90, mirror_x, mirror_y Currently RED on rotation 90/270 (rotation direction) and mirror_x/y (axis semantics swapped). Production fix in WireDragger.pin_world_xy to follow. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	2026-05-03 22:24:14 +01:00

Eugene Mikhantyev

496be317e9

test: close mirror-fixture gap exposed by post-fix audit

The eeschema-ground-truth fixture (_build_mirror_case) passed
'mirror': 'x' to ComponentManager.add_component, which silently drops
the kwarg — so the resulting .kicad_sch had no (mirror x|y) token,
eeschema rendered an unmirrored symbol, and our pin coords (also
unmirrored) tautologically matched. The mirror tests were GREEN both
before and after the rotation/mirror fix in 7e67cb9, providing zero
regression coverage for the mirror semantics.

Fix:
  - _build_mirror_case now applies the mirror via the same low-level
    helper (WireDragger.update_symbol_rotation_mirror) that
    rotate_schematic_component uses, with a guard assertion that the
    written file actually contains (mirror x|y).
  - Two new pin-down unit tests in test_add_schematic_component.py
    document and lock down ComponentManager.add_component's silent-drop
    behavior for mirror, so the next person to touch that path knows to
    update the eeschema-truth fixture if they grow real mirror support.

Verified: with the production fix at 7e67cb9 reverted, the kicad-cli
mirror tests now go RED (previously they stayed GREEN regardless).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-05-03 22:40:34 +01:00

Eugene Mikhantyev

f7660e15ad

test: encode eeschema kicad-cli ground truth (currently RED)

Update existing rotation/mirror assertions and add a kicad-cli-grounded
regression suite. eeschema's pin transform is verified against:
  - rotation: TRANSFORM(0,1,-1,0) for rot=90 — CCW in screen Y-down
  - mirror x: SYM_MIRROR_X = TRANSFORM(1,0,0,-1) — negates internal Y
  - mirror y: SYM_MIRROR_Y = TRANSFORM(-1,0,0,1) — negates internal X

Updated assertions:
  - test_pin_locator_y_flip: rotated cap pin 1 X 153.81 → 146.19
  - test_move_with_wire_preservation::test_resistor_rotated_90:
    pin 1 X 103.81 → 96.19
  - test_hierarchical_pad_net_map::test_90_degree_rotation:
    swapped UP_NET / DN_NET label coords

New file tests/test_pin_world_xy_eeschema_truth.py:
  - 4 parametrized diode tests via kicad-cli netlist (oracle = eeschema)
  - 2 parametrized resistor mirror tests via kicad-cli netlist
  - 3 pure-math pin_world_xy assertions for rot=90, mirror_x, mirror_y

Currently RED on rotation 90/270 (rotation direction) and mirror_x/y
(axis semantics swapped). Production fix in WireDragger.pin_world_xy
to follow.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-05-03 22:24:14 +01:00

2 Commits