* feat: guard SWIG auto-save against external file changes
After every board-mutating SWIG command, kicad_interface._auto_save_board()
unconditionally calls pcbnew.SaveBoard() with the in-memory board. When the
on-disk .kicad_pcb has been modified externally between our LoadBoard and
SaveBoard (KiCad GUI's own save, git checkout, another process), the
in-memory state silently overwrites those external changes - losing data
the user can't see was at risk.
This change records the file's mtime_ns + sha256 at LoadBoard and verifies
the signature matches before each auto-save:
* If the signature has diverged, refuse the save and attach a structured
warning to the command result so callers know their mutation is
in-memory only and they need to reload before retrying.
* If it matches, copy the existing file to .mcp-backups/<name>.<ts>
(rotating, keeps last 20) before overwriting.
* Update the recorded signature after our own writes so subsequent
saves are not falsely flagged.
Backwards compatible:
* No tool schemas changed.
* Successful saves return as before, with an extra `autoSave` field
when the wrapper observed something noteworthy.
* Refused saves return success: true (the in-memory mutation did
succeed) plus warnings: [...] and autoSave.diskChangedExternally,
so callers can detect the situation programmatically.
Adds tests/test_auto_save_guard.py (10 tests, all passing) covering:
signature math, refusal on external change, backup creation + content,
backup rotation, first-save semantics (no recorded signature proceeds
normally), and skip cases (no board / no path).
Motivation: the aircam-pdb fork-user lost ~480 traces and the full
footprint layout to a silent overwrite incident on 2026-05-03; recovery
was only possible because VS Code's local-history extension happened to
have a snapshot from a few minutes earlier. This guard makes that class
of incident loud and locally recoverable.
* fix(auto-save-guard): refuse only on content divergence, not mtime
The guard added in 9ba0010 records `(mtime_ns, sha256)` as the file's
disk signature and refuses auto-save when the recorded tuple no longer
matches the current one. Comparing the full tuple meant any mtime delta
fired the refusal — including a bare `touch` of the file, an atime-style
backup tool, or any MCP read path that opened the .kicad_pcb between
load and save. Users were trapped: every write needed an explicit
save_project call to bypass the false positive (documented as a
workaround in fork users' notes).
Compare on sha256 only; mtime is incidental. The actual data-loss
scenario the guard is meant to catch — an external write that genuinely
changed the file — produces a different content hash, which is what the
guard now keys off. After a touch-only mtime advance with content
unchanged, refresh the recorded signature so we don't re-hash on every
subsequent call.
Drops the mtime-equality fast path on _disk_signature: a filesystem with
coarse mtime resolution (FAT32, some network mounts) could accept two
writes inside one mtime tick; trusting mtime as a hash cache key would
re-introduce a class of silent overwrites the guard exists to prevent.
The hash itself is cheap (sha256 over a typical .kicad_pcb completes in
tens of milliseconds).
Adds 4 regression tests in test_auto_save_guard.py:
- touch-only mtime advance proceeds and refreshes the signature
- content change at the same mtime is still refused (hash divergence
must drive the decision, not tuple equality)
- the user-facing warning calls out "contents", not "mtime"
- _disk_signature returns the same hash when content is unchanged
even after the file's mtime advances
The handler iterated `board.GetFootprints()` and assigned nets to existing
pads, but had no path to *add* footprints for schematic symbols whose
Reference was not yet on the board. New parts placed on the schematic
landed in the net list with no PCB representation — the rats nest had
nowhere to terminate, and place_component / route_pad_to_pad would fail
because the target footprint did not exist.
KiCad's "Update PCB from Schematic" (F8) implicitly adds the missing
footprints; bring the MCP's behaviour in line.
Implementation:
* `_extract_components_from_schematic` runs `kicad-cli sch export netlist
--format kicadxml` (the same path `_handle_generate_netlist` already
uses) and returns a flat `[{reference, value, footprint}]` list. Walks
hierarchical sub-sheets transparently because kicad-cli does.
* `_add_missing_footprints_from_schematic` resolves each missing component
against the project's fp-lib-table via `LibraryManager`, calls
`pcbnew.FootprintLoad`, sets reference / value / FPID, and places the
footprint at the board origin (the user / autoplacer can position it).
Power and flag references (`#PWR…`, `#FLG…`) are excluded — they have
no PCB footprint.
* The pad-net assignment loop now runs *after* the add path, so newly
placed footprints get their nets assigned in the same call.
* Response payload gains `footprints_added` and `footprints_skipped`
diagnostic lists. The textual `message` field reports both the new
footprint count and the existing net / pad counts.
Adds tests/test_sync_schematic_to_board_footprints.py — 9 unit tests
covering the add path (missing ref, already-present ref, power refs,
empty footprint, unknown library) and the kicad-cli helper (XML parse,
missing kicad-cli, non-zero exit).
WireManager.add_wire and add_polyline_wire bailed out with
"No sheet_instances section found in schematic" when called on a
hierarchical sub-sheet — that block only exists in the root .kicad_sch.
The handler in kicad_interface.py converted the False return into a
flat "Failed to add wire" message, leaving callers with no diagnostic.
Apply the same fallback that WireManager.add_label has used since the
hierarchical-design support: when (sheet_instances ...) is absent,
append the new (wire ...) item at the end of the outer (kicad_sch ...)
form.
Adds tests/test_add_wire_sub_sheet.py with 6 regression tests covering
both add_wire and add_polyline_wire on a sub-sheet, including paren
balance, sexpdata round-trip, and segment count for an N-point polyline.
DynamicSymbolLoader.create_component_instance only handled root
schematics: it located its insertion point via `(sheet_instances`,
which exists only in the root .kicad_sch. Adding a component to any
hierarchical sub-sheet raised "Could not find insertion point in
schematic" and aborted the call.
Fall back to inserting just before the closing paren of the outer
(kicad_sch ...) form when the marker is absent. WireManager.add_label
already uses the same fallback for hierarchical sub-sheets.
Adds three regression tests in TestCreateComponentInstanceSubSheet
covering successful insertion, paren balance, and sexpdata round-trip
on a sub-sheet without (sheet_instances).
`BoardOutlineCommands.add_mounting_hole` produced footprints with an empty
library:name id (`(footprint "" ...)` in the .kicad_pcb), which KiCad's GUI
Move tool refuses to select — users couldn't drag the resulting MHs in the
editor. It also emitted the pad with the default `*.Cu` + `*.Mask` LSET on
NPTHs; with `padDiameter > diameter` that creates phantom copper annular
rings on every Cu layer and trips clearance DRC against neighbouring nets.
Repro: call `add_mounting_hole` with `position={x:117,y:84.5,unit:"mm"},
diameter:3.2, padDiameter:3.5`. The resulting MH is unselectable and DRC
reports phantom F.Cu pad shorts to neighbouring component pads.
Changes:
- Set a real FPID via `module.SetFPID(pcbnew.LIB_ID(lib, name))`. Default
is `MountingHole:MountingHole_<diameter:g>mm` (e.g. 3.2 → 3.2mm); a new
optional `footprintLibId` parameter overrides.
- For NPTH (the default `plated:false`), restrict the pad's LSET to
`F.Mask` + `B.Mask` only. PTH path is unchanged — the default Cu+Mask
LSET is correct there.
- Update the schema in `tool_schemas.py`: previously advertised
`x`/`y`/`diameter` at the top level, but the impl reads
`position={x,y,unit}`, `padDiameter`, `plated`. Schema now matches the
implementation and exposes the new `footprintLibId` param.
- New `tests/test_add_mounting_hole.py` regression suite (7 tests) asserting
SetFPID is invoked with non-empty lib:name (default + override forms),
NPTH SetLayerSet excludes any Cu layer, and PTH does not call
SetLayerSet (preserves default Cu+Mask).
Some community-generated KiCad symbol libraries (e.g.
PCM_Diode_Schottky_AKL:MBRS130) define each pin number twice — once as a
visible "real" pin with non-zero length whose ``at`` coordinate is the
wire-connection endpoint, and once as an inner zero-length "ghost" pin
used as an internal graphic-anchoring join. Both definitions live inside
the same ``lib_symbols`` block.
``parse_symbol_definition`` stored pins via ``pins[number] = pin_data``
— a plain assignment. Each duplicate-numbered pin encountered during the
recursive walk overwrote the previous one. The recursion order put the
ghost pins last for MBRS130, so the ghost won and ``get_pin_location``
returned a coordinate that did not match any wire/label.
Downstream this caused ``get_connections_for_net`` to silently miss diode
pins on the rails they were wired to — on a real schematic, querying
``+BATT`` returned 8 of 9 expected nodes (D1/1 absent) and ``+3V3``
returned 44 of 46 (D1/2 and D2/2 absent), because the BFS could not find
the diode's pin endpoint at the labelled position.
Fix: when the same pin number is defined more than once, keep the entry
with the greater ``length``. The outer real pin has length > 0; the
inner ghost has length == 0. Strict-greater comparison resolves ties to
first-encountered, so legitimate same-length duplicates (e.g., per-unit
repetitions in multi-unit symbols) keep stable existing behaviour.
Tests: four unit tests in ``tests/test_pin_locator_duplicate_pin_defs.py``
cover (a) outer-then-ghost order (the real bug), (b) ghost-then-outer
order (length-not-order heuristic), (c) no-duplicate baseline regression,
and (d) equal-length tie keeps first-encountered. Two of the four fail
on main, all four pass on this branch.
Full suite: 671 passed, 11 skipped, 0 regressions (modulo the pre-existing
tests/test_get_pin_angle.py collection error which is unrelated to
pin_locator).
End-to-end on the real schematic that triggered the report: after the
fix ``get_connections_for_net('+BATT')`` returns all 9 expected nodes
matching ``kicad-cli sch export netlist`` exactly. The companion fix in
PR #177 (wire_connectivity pwr-flag bridge) closes the orthogonal
over-merge bug; together they bring net membership to full parity with
kicad-cli on schematics that use both ``PWR_FLAG`` markers and
diodes from community libraries.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Every power:PWR_FLAG symbol carries Value="PWR_FLAG" — the ERC marker
inherits its rail's name from the wire/label it sits on. Commit 7f3a379
added #FLG symbols to the same handling loop as #PWR power ports inside
_parse_virtual_connections, so every pwr-flag was getting appended to
label_to_points["PWR_FLAG"]. The BFS in _find_connected_wires uses
label_to_points for virtual jumps; reaching any pwr-flag pin caused it
to teleport to every other pwr-flag pin, walking across each one's stub
wire into a different power rail. The result: get_net_connections(rail)
returned the union of pins on every rail that had a pwr-flag, for any
rail.
Fix: pwr-flag pin positions still register as anchors in point_to_label
(preserving the original intent of 7f3a379 so find_orphaned_wires keeps
accepting them), but they no longer enter label_to_points. The pwr-flag
remains electrically connected to its rail via the wire-graph BFS through
the wire it sits on; the label-jump mechanism is unnecessary for that
path and actively harmful when the "label" is the same for unrelated
rails.
Tests: three unit tests on _parse_virtual_connections cover the bug
(over-merge gone), regression check (power ports still work in both maps),
and edge case (pwr-flag and port at same point — port name wins). All
three fail on main, pass on this branch.
Full suite: 667 passed, 11 skipped, 0 regressions (modulo the pre-existing
tests/test_get_pin_angle.py collection error which is unrelated to
wire_connectivity).
End-to-end verification on a single-sheet schematic with 7 distinct
power rails each carrying a pwr-flag: every queried net now matches
the official kicad-cli netlist output (modulo a separate library-symbol
bug on PCM_Diode_Schottky_AKL:MBRS130 with duplicate pin definitions,
out of scope here).
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
dynamic_symbol_loader only consulted the project-local sym-lib-table
and a hardcoded list of bundled symbol directories, so libraries
registered via the user-global sym-lib-table (Preferences > Manage
Symbol Libraries > Global) were invisible to add_schematic_component.
Common case: company libraries that live under OneDrive / a network
share / any other custom path the user added through the GUI.
Also widened the sym-lib-table parser regex to accept quoted URIs (and
quoted names) that contain spaces — required for paths like
"C:/Users/.../OneDrive - Company/Documents/KiCad/...". The old bare-
word capture stopped at the first space.
Search order is now:
1. Project sym-lib-table
2. User-global sym-lib-table (~/AppData/Roaming/kicad/<ver>/sym-lib-table
on Windows, ~/.config/kicad/<ver>/sym-lib-table on Linux,
~/Library/Preferences/kicad/<ver>/sym-lib-table on macOS)
3. Bundled / well-known symbol directories
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
cairocffi uses cffi's ffi.dlopen('cairo-2') which requires the DLL to
be on the system PATH. On Windows, prepend KiCad's bin directory to
PATH early in kicad_interface.py (before any cairocffi import) so
cairo-2.dll can be found. Checks PYTHONPATH, Python executable dir,
and default KiCad 9/8 install paths.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The _ipc_move_component and _ipc_place_component handlers were ignoring
the unit field from position parameters, always treating values as mm.
When inches were specified, components would be placed at 1/25.4th of
the intended position. Now reads the unit field and converts to mm
before passing to the IPC backend.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
KiCad 9's IPC API returns Footprint objects where fp.definition lacks
the library_link attribute, causing all component queries to fail with
"'Footprint' object has no attribute 'library_link'" warnings and
returning 0 components. This adds a hasattr check with fallback to
fp.definition.id.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two bugs in WireDragger.pin_world_xy (and corresponding bugs in
PinLocator.get_pin_angle) caused pin coordinates and angles to land on
the wrong pin in 4 of 8 polarized cases (rot=90, rot=270, mirror x on a
vertical part, mirror y on a vertical part). Verified end-to-end against
`kicad-cli sch export netlist`.
(1) Rotation direction. After PR #145's `-ly` Y-flip, calling the
standard math (Y-up CCW) `_rotate` is effectively CW in screen Y-down.
eeschema's TRANSFORM(0,1,-1,0) for rot=90 is screen-CCW. They agreed at
0° and 180° (where the rotation matrices coincide) but disagreed at 90°
and 270°.
(2) Mirror axis semantics swapped. Per eeschema symbol.h:43-44,
SYM_MIRROR_X = TRANSFORM(1,0,0,-1) negates Y, and SYM_MIRROR_Y =
TRANSFORM(-1,0,0,1) negates X. Our code did the inverse: `mirror_x`
negated the X component and `mirror_y` negated the Y component.
Fix shape for `_rotate`: chose option (b) — leave `_rotate` as standard
math and negate the angle at the call site (`_rotate(lx, ly, -rotation)`).
This converts math-CCW to screen-CCW without disturbing
`TestRotatePoint`'s direct expectations of `_rotate`.
Final composition order in `pin_world_xy` matches eeschema's parser
(rotation set first into m_transform, then mirror composed via
`new = old * temp` so the mirror is applied first to the coordinate):
1. Y-flip: ly = -ly (lib Y-up → screen Y-down)
2. Mirror: if mirror_x: ly = -ly (negate screen-Y)
if mirror_y: lx = -lx (negate screen-X)
3. Rotate: _rotate(lx, ly, -rotation) (screen-CCW)
4. Translate: add (sym_x, sym_y)
Verified by hand for {rot=90, rot=270} × {none, mirror_x, mirror_y}
against the TRANSFORM matrices in transform.cpp:44 and symbol.h:43-44.
`PinLocator.get_pin_angle` mirrors the same composition in angle space.
For an angle, Y-flip and mirror_x both negate the angle; mirror_y maps
to (180 - angle). The screen-CCW rotation in `pin_world_xy` corresponds
to subtracting (not adding) the symbol rotation in standard atan2
convention — fixed accordingly. Geometry test
(`test_get_pin_angle.py::test_get_pin_angle_matches_geometric_expectation`)
derives expected angles from `pin_world_xy` itself, so it pins the two
together.
`tests/test_rotate_schematic_mirror.py::test_pin_positions_mirror_x_flips_x`
encoded the OLD inverted semantics and is updated/renamed to
`test_pin_positions_mirror_x_flips_y` with a pin that has non-zero Y so
the assertion is meaningful under the corrected semantics.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Resolves the four failing tests in tests/test_pin_locator_and_component.py
left behind by the PR #145 / commit 3c22580 Y-flip work.
Per-test rationale:
- TestPinLocatorYAxisNegation::{test_pin1_y_above_center_for_rotation_0,
test_pin2_y_below_center_for_rotation_0, test_pin1_rotated_90}: stale.
Their assertions encoded the *correct* post-PR-145 convention (96.19,
103.81, etc.), but their setup MagicMock'd self._schematic_cache while
bypassing _get_symbol_transform, which reads the .kicad_sch file
directly via sexpdata. The end-to-end Y-flip behaviour is already
covered against eeschema in tests/test_pin_locator_y_flip.py — keeping
three mock-based duplicates added no value, so they were removed.
- TestPinLocatorReferenceRstrip::test_get_pin_location_finds_symbol_with_trailing_underscore:
revealed a real production bug. PinLocator.get_pin_location strips a
trailing "_" on the kicad-skip lookup path, but the sexpdata-based
_get_symbol_transform delegates to WireDragger.find_symbol which used an
exact-equality comparison. With kicad-skip's "R1_" artifact the function
returned None, so the whole pin-location call failed even when the symbol
was clearly present. Fixed find_symbol to apply the same rstrip("_") on
the stored reference before comparing, mirroring the existing behaviour
in PinLocator. The test was also rewritten to use a real temp .kicad_sch
(with the on-disk reference mangled to "R1_") so it actually exercises
both lookup paths instead of bypassing one with mocks.
Files changed:
- python/commands/wire_dragger.py:78-89 — rstrip("_") on the reference
read out of the symbol property before comparing to the caller-supplied
reference.
- tests/test_pin_locator_and_component.py — removed three stale mock-based
Y-axis tests (covered by tests/test_pin_locator_y_flip.py end-to-end);
rewrote rstrip tests to use a real schematic file so _get_symbol_transform
is actually exercised.
Verified: tests/test_pin_locator_and_component.py + test_pin_locator_y_flip.py
+ test_get_pin_angle.py + test_move_with_wire_preservation.py — 69 passed.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
PR #145 restored the Y-axis flip in WireDragger.pin_world_xy so pin
coordinates now match the schematic (Y-down) frame instead of the
library (Y-up) frame. PinLocator.get_pin_angle was the companion to
that transform but never received the matching fix: it was returning
the library-frame angle (with mirror handling but no Y-flip), so
angles came out 180° off along the Y axis.
This was masked before PR #145 because pin_world_xy was wrong in the
same direction — both functions skipped the Y-flip, so callers that
compared pin endpoints to angles saw a self-consistent picture. Once
pin_world_xy was corrected the inconsistency surfaced.
Apply the same lib→screen Y-flip (negate angle) after the mirror
handling and before the symbol-rotation add, matching pin_world_xy's
order: mirror in lib space → Y-flip → rotate → translate (no
translate for angles since angles are translation-invariant).
Fixes the 24 parametrized cases in
tests/test_get_pin_angle.py::test_get_pin_angle_matches_geometric_expectation
(pin × mirror × rotation matrix). The test derives its expected value
from pin_world_xy itself, making it the canonical geometric oracle.
test_pin_locator_y_flip and test_move_with_wire_preservation continue
to pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
After every board-mutating SWIG command, kicad_interface._auto_save_board()
unconditionally calls pcbnew.SaveBoard() with the in-memory board. When the
on-disk .kicad_pcb has been modified externally between our LoadBoard and
SaveBoard (KiCad GUI's own save, git checkout, another process), the
in-memory state silently overwrites those external changes - losing data
the user can't see was at risk.
This change records the file's mtime_ns + sha256 at LoadBoard and verifies
the signature matches before each auto-save:
* If the signature has diverged, refuse the save and attach a structured
warning to the command result so callers know their mutation is
in-memory only and they need to reload before retrying.
* If it matches, copy the existing file to .mcp-backups/<name>.<ts>
(rotating, keeps last 20) before overwriting.
* Update the recorded signature after our own writes so subsequent
saves are not falsely flagged.
Backwards compatible:
* No tool schemas changed.
* Successful saves return as before, with an extra `autoSave` field
when the wrapper observed something noteworthy.
* Refused saves return success: true (the in-memory mutation did
succeed) plus warnings: [...] and autoSave.diskChangedExternally,
so callers can detect the situation programmatically.
Adds tests/test_auto_save_guard.py (10 tests, all passing) covering:
signature math, refusal on external change, backup creation + content,
backup rotation, first-save semantics (no recorded signature proceeds
normally), and skip cases (no board / no path).
Motivation: the aircam-pdb fork-user lost ~480 traces and the full
footprint layout to a silent overwrite incident on 2026-05-03; recovery
was only possible because VS Code's local-history extension happened to
have a snapshot from a few minutes earlier. This guard makes that class
of incident loud and locally recoverable.
search_symbols, list_symbol_libraries, list_library_symbols, and
get_symbol_info previously only consulted the global sym-lib-table. A
library registered with project scope (an entry in
<project>/sym-lib-table) was therefore invisible — even right after
open_project succeeded — making add_schematic_component the only tool
that could see it.
Fix has two parts:
1. Wrap project_commands.open_project and project_commands.create_project
in handlers that rebuild SymbolLibraryCommands.library_manager against
the project directory. After open_project, project-scope libraries are
automatically visible to subsequent search/list/info calls.
2. Add an optional projectPath parameter to the four discovery tools
(accepts a project directory, .kicad_pro, .kicad_pcb, or .kicad_sch
path). Stateless callers can resolve project libraries without first
calling open_project. SymbolLibraryCommands._derive_project_path also
walks up from schematicPath/boardPath to find the directory that owns
the project, mirroring the logic in _handle_add_schematic_component.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The KiCad s-expression parser rejects raw newline and carriage-return
characters inside quoted string literals — a multi-line text annotation
written through `add_text` produced a `.kicad_sch` file that eeschema
silently tolerated but `kicad-cli sch ...` refused with "Failed to load
schematic." The escape pass only handled backslashes and double quotes.
Add `\\n` → `\\\\n` and `\\r` → `\\\\r` to the same escape chain. Order
matters: backslashes are escaped first so we don't double-escape our
own escapes.
A new regression test (`test_escapes_newlines_in_multiline_text`)
checks both that the resulting quoted string literal contains no raw
newline characters and that the file round-trips cleanly through the
sexpdata parser.
End-to-end smoke: a 4-line annotation written through the patched
add_text now passes `kicad-cli sch erc` (exit 0) where the previous
behaviour failed parse.
Note: the same escape gap exists in `_make_hierarchical_label_text`
and `_make_sheet_pin_text` for unescaped quotes/newlines in the user-
supplied text. Not fixed here to keep this PR scoped to the documented
add_text bug; happy to fold it in if a reviewer prefers.
Library symbol pins are stored Y-up (positive Y is upward in the symbol
editor's coordinate system) but `.kicad_sch` is Y-down (positive Y is
downward in the schematic). `pin_world_xy` was returning `sym_y + ry`
without negating the rotated lib Y, so for any non-symmetric symbol
pin 1 and pin 2 ended up at swapped world positions.
For symmetric two-pin passives (R, C non-polarized) this was invisible
because pin 1 and pin 2 are electrically equivalent. For polarized
parts — electrolytic and polymer caps, diodes, MOSFETs, BJTs — it
silently swapped polarity. A label snapped to a polarized cap's pin 1
ended up on pin 2, which is catastrophic at first power-up.
The order matches eeschema's actual transformation:
mirror in lib space → Y-flip to screen → rotate → translate.
The existing regression test in test_pin_locator_y_flip.py was already
written with the correct expected coordinates but the matching code fix
was never landed; that test now passes.
Three tests in test_move_with_wire_preservation.py had baked the buggy
expected coordinates into their assertions; updated those to the correct
y-flipped values. The touching-pin fixture had to flip R2's Y from
-7.62 to +7.62 so the two pins still meet under the corrected formula.
Verified end-to-end on a 46-component aerospace PDB schematic: all
8 polarized-part pins (4 polymer caps + 4 TVS diodes) now produce
world coordinates that match the labels actually placed in the file.
1. add_schematic_net_label failed on schematics with no existing labels.
The clone-based path required a pre-existing label to copy from;
the documented "fallback to sexpdata" was a misleading log line —
the RuntimeError was caught and the call silently returned False.
Restore hand-built sexpdata construction (without the buggy
fields_autoplaced token, with orientation-aware justify).
2. get_pin_angle returned the wrong angle for every mirrored symbol
(off by exactly 180°, all rotations, both mirror axes). The
mirror_x and mirror_y formulas were swapped relative to the
pin_world_xy convention — pin_world_xy mirrors a position by
flipping its local axis component, so the matching angle
transform is (180 - θ) for mirror_x and -θ for mirror_y.
Add regression tests:
- test_add_label_empty_schematic.py — first label on empty schematic,
orientation-aware justify.
- test_get_pin_angle.py — full 24-case matrix
(4 rotations × 3 mirror states × 2 pins).
Previously get_pin_location and get_pin_angle read symbol state from a
kicad-skip cache that does not reflect (mirror x/y) tokens written by
rotate_schematic_component. Pin coordinates were always computed as if
the symbol was unmirrored.
Fix:
- Added _get_symbol_transform() which reads position, rotation, mirror_x,
mirror_y, and lib_id directly from the .kicad_sch file via sexpdata +
WireDragger.find_symbol (the authoritative source after a rotate/mirror)
- get_pin_location now delegates the full transform (mirror → rotate →
translate) to WireDragger.pin_world_xy, matching the logic used by
move_schematic_component and rotate_schematic_component
- get_pin_angle now applies mirror-induced angle reflection before adding
symbol rotation: mirror_x negates the angle, mirror_y reflects across 180°
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two bugs fixed:
1. fields_autoplaced yes was always injected — caused incorrect visual
rendering of label text in KiCAD. Removed by using clone() which
copies an existing label without that field.
2. (justify left bottom) was hardcoded regardless of orientation.
For orientation 180/270 KiCAD requires (justify right bottom).
Now set correctly via new_label.effects.justify._tree[1].
Implementation switches from manual sexpdata list construction to
kicad-skip Schematic.label[0].clone(), which produces a structurally
correct label that KiCAD can round-trip without modification.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previously the handler used kicad-skip to apply rotation and mirror.
kicad-skip has no API for (mirror x/y) on placed symbols, causing:
'NoneType' object has no attribute 'value'
Fix:
- Rewrote _handle_rotate_schematic_component to use sexpdata (same
approach as move_schematic_component) for both rotation and mirror
- Added WireDragger.compute_pin_positions_for_rotation: computes old
and new pin world positions when rotation/mirror changes at fixed (x,y)
- Added WireDragger.update_symbol_rotation_mirror: updates (at) rotation
and adds/removes/replaces the (mirror x/y) sexpdata token cleanly
- Connected wires now follow pin positions after rotate/mirror via the
existing WireDragger.drag_wires infrastructure
Tests: 10 unit tests in tests/test_rotate_schematic_mirror.py covering
update_symbol_rotation_mirror, compute_pin_positions_for_rotation, and
a handler smoke test.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
PWR_FLAG instances use a #FLG reference prefix, not #PWR, so their
pin positions were never registered as virtual connection anchors in
_parse_virtual_connections. As a result, find_orphaned_wires reported
wire ends terminating on a PWR_FLAG as dangling. Other call sites
(schematic_analysis.py:127, kicad_interface.py:3814) already recognize
#FLG as a power symbol; align this site with them.
Also coerce a previously-validated Optional[int] to int in board/layers.py
so the file passes mypy (required by the pre-commit hook); behavior is
unchanged because the value is already None-checked above.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Refactor _handle_rotate_schematic_component to use raw sexp throughout
and write the schematic once instead of three times
- Hoist sexpdata and WireManager imports to module scope; drop the
unnecessary underscore aliases in move/rotate handlers
- Move WireManager._SUB_UNIT_RE to the top of the class body
- Promote the per-call Symbol("symbol")/Symbol("unit") allocations in
_parse_lib_pins / _collect_pin_positions to module-level _SYM_*
constants
- Document the assumption behind _SUB_UNIT_RE's greedy match
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the manual add_schematic_junction tool with automatic junction
management. WireManager.sync_junctions inserts/removes junction dots
based on wire endpoints plus component pin positions and is invoked
after add_wire, add_polyline_wire, delete_wire, move, and rotate.
- Pin-aware: parses lib_symbols and applies KiCad's mirror/rotate/
translate transform to compute world pin coordinates
- Multi-unit safe: filters lib_symbols sub-units by the placed
symbol's (unit N) field plus the unit-0 common body
- Removes the now-unused WireManager.add_junction static method
- Updates CHANGELOG [Unreleased] with the tool removal notice
- Adds .mcp.json to .gitignore (machine-local paths)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Configuring logging at module import time hard-failed in environments
without write access to the user's home directory (sandboxed pytest
runners, restricted CI containers, read-only filesystems). The
unhandled OSError/PermissionError aborted module import and cascaded
into ~100 test failures during collection.
Wrap the FileHandler setup in try/except and fall back to console-only
logging when the log directory cannot be created. Production behavior
is unchanged - file logging in ~/.kicad-mcp/logs is still used
whenever the directory is writable.
Also picks up an isort-driven import reorder applied by the project's
pre-commit hook.
The symbol-to-schematic y-flip was applied twice in sequence (two identical
negation blocks with matching comments), cancelling out and leaving pin
Y-coordinates mirrored about the placement Y. For symmetric passives the
bug is invisible (pin 1 and pin 2 are electrically interchangeable); for
ICs with non-equivalent pins (power pins, opamp inputs, etc.) this causes
tools that go through PinLocator — connect_to_net, add_schematic_connection,
add_schematic_net_label with componentRef+pinNumber — to place connections
at the mirror-flipped pin. Verified against kicad-cli generate_netlist
ground truth: on a Device:R placed at (111.76, 83.82), pin 1 resolves to
y=80.01 (actual) vs y=87.63 (pre-fix).
This is a regression of PR #103, which originally fixed the y-negation;
the redundant second block was added subsequently.
Includes a regression test with both a straight Device:R and a rotated
Device:C to exercise the y-flip + rotation pipeline.
connect_to_net and connect_passthrough previously only touched the
schematic: a wire stub plus a net label were added via
ConnectionManager, but the corresponding pad on the .kicad_pcb was
never updated.
pcbnew.SaveBoard() silently drops every net that is not referenced by
at least one board element (pad/track/via/zone). A net that exists
only in the schematic therefore disappears on save, and in KiCad the
board's ratsnest never shows anything.
Fix:
- new _assign_net_to_pad(component_ref, pin_name, net_name):
ensures the net exists on the board (creates NETINFO_ITEM if not)
and sets it on the matching pad of the matching footprint.
- _handle_connect_to_net calls it after the schematic op when
self.board is loaded.
- _handle_connect_passthrough parses ConnectionManager's 'connected'
entries ('J1/1 <-> J2/1 [PIN_1]') and assigns nets to both pads.
- register 'connect_to_net' in _BOARD_MUTATING_COMMANDS so the
existing auto-save path persists the pad assignment to disk.
Passing net_name='*' now deletes all tracks on the board (respecting
the include_vias flag). Useful for wiping a test layout before
re-routing without having to iterate every net by name.
board.GetLayerStack() was removed in KiCad 9. Call SetLayerName and
SetLayerType directly on the board instead, and grow the copper layer
count via SetCopperLayerCount when adding inner layers. Without this,
add_layer raises AttributeError on any KiCad 9 installation.
KiCad 10.0 on Windows ships pcbnew.py under bin/Lib/site-packages rather
than the lib/python3/dist-packages path used by KiCad 9.x. Without this
path the MCP server's Python process fails to import pcbnew and exits
immediately with "Failed to import pcbnew module - KiCAD Python API not found".
Changes:
- Prioritise 10.0 in the version list so the newer path is checked first
- Add bin/Lib/site-packages check before the existing lib/python3/dist-packages
check for every version, so both layouts are supported side-by-side
Note: the pcbnew extension is compiled against KiCad's bundled Python 3.11,
so callers must also set KICAD_PYTHON to the KiCad python.exe
(C:\Program Files\KiCad\10.0\bin\python.exe) to avoid a DLL version
conflict when the system Python differs.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
KiCad serialises rescued or locally-customised library entries with an
extra (lib_name "...") child before (lib_id "..."):
(symbol
(lib_name "RESISTOR_0603_4")
(lib_id "MF_Passives:RESISTOR_0603")
(at 132.08 44.45 90)
...)
The block-matching regex in _handle_get_schematic_component,
_handle_edit_schematic_component, and _handle_delete_schematic_component
required (lib_id IMMEDIATELY after (symbol, so any placed component
using this form was silently invisible to lookup. The user-visible
symptom is "Component '<ref>' not found in schematic" even though the
component is plainly present (and reachable through list / IPC paths).
This bug also affected set/remove_schematic_component_property and the
existing footprint/value/reference rewriting paths in edit, since they
all share the same lookup code.
The parent-position lookup used a similarly-strict regex
((symbol (lib_id "...") (at ...))), which silently fell back to (0,0)
on (lib_name)-first symbols and caused new properties added through
the custom-properties path to anchor at the schematic origin instead
of the parent symbol.
Fix: relax the symbol-block opening pattern to (symbol\s+\( — matching
any opening paren after (symbol — and read the symbol's origin from
the first (at ...) inside the block. Library-definition entries inside
(lib_symbols ...) are still excluded by the existing range check
(they use the (symbol "name" ...) form with a quoted string, not a
paren).
Adds 7 regression tests in TestLibNameBeforeLibIdOrdering using a
real-world (lib_name)-first resistor block, covering get / edit /
set-property / remove-property / delete and verifying that newly
added properties anchor to the symbol origin instead of (0, 0).
