fix(templates): name symbols and pins that exist in KiCad 10 (#783)

[YORROY123]

Summary

Two bundled templates could not be applied on KiCad 10 at all, and four templates sent the agent to pins their symbols do not have. This fixes every template's symbol and pin names, gives ldo_3v3 and stm32_minimal the content #783 asks for, and adds a test that checks the templates against the installed KiCad library.

Closes #783

Measured on main @ 9488e5f, applying each bundled template to a fresh copy of tests/fixtures/derived_lib_name.kicad_sch (KiCad 10.0.6, Windows 11):

template	before	after
usb_c_5v_sink	refused: Connector:USB_C_Receptacle_USB2.0 not found	placed
stm32_minimal	refused: Connector:Conn_ARM_SWD_10 not found	placed
ldo_3v3, i2c_pullups, led_indicator, buck_converter	placed	placed

kicad-cli sch export netlist loads every result; the stm32_minimal netlist lists all 15 template parts with their pins.

The new test, run against the unchanged templates, reports:

usb_c_5v_sink: Connector:USB_C_Receptacle_USB2.0 is not in the KiCad library
ldo_3v3: U.VIN — Regulator_Linear:AMS1117-3.3 has no pin VIN
ldo_3v3: U.VOUT — Regulator_Linear:AMS1117-3.3 has no pin VOUT
stm32_minimal: Connector:Conn_ARM_SWD_10 is not in the KiCad library
stm32_minimal: U.OSC_IN — MCU_ST_STM32F4:STM32F411CEUx has no pin OSC_IN
stm32_minimal: U.OSC_OUT — MCU_ST_STM32F4:STM32F411CEUx has no pin OSC_OUT
stm32_minimal: U.SWDIO — MCU_ST_STM32F4:STM32F411CEUx has no pin SWDIO
stm32_minimal: U.SWCLK — MCU_ST_STM32F4:STM32F411CEUx has no pin SWCLK
buck_converter: U.FB — Regulator_Switching:TPS563200 has no pin FB
buck_converter: U.BOOT — Regulator_Switching:TPS563200 has no pin BOOT

Three of these were not listed in #783: the USB-C symbol name and the TPS563200 FB/BOOT pins (VFB/VBST).

Approach

Symbol and pin names. USB_C_Receptacle_USB2.0 → USB_C_Receptacle_USB2.0_16P; Conn_ARM_SWD_10 → Conn_ARM_JTAG_SWD_10; STM32 pins by name (PH0, PH1, PA13, PA14); U.FB/U.BOOT → U.VFB/U.VBST.

ldo_3v3. AMS1117 with a 10 µF ceramic output contradicted the template's own note. It now uses AP2112K-3.3, as #783 suggests. The values below come from the Diodes datasheet (DS39724 Rev. 2-2):

• It is stable with 1 µF ceramic capacitors on input and output, so the template lists two 1 µF capacitors.
• EN has an internal 3 MΩ pull-down, so it is mapped to VIN. Left open, the regulator stays off.
• The design notes give the 6.5 V absolute maximum input. On a hot-plugged USB input, cable inductance rings with a small input capacitor and can overshoot that limit.
• Dropout is 250 mV typical and 400 mV max at 600 mA.

stm32_minimal. Every item in #783's acceptance list:

• VCAP1 4.7 µF, ESR < 1 Ω
• BOOT0 and PB2 each pulled down with 10 kΩ
• VBAT, VSSA and VCAP1 mapped
• four 100 nF capacitors: one per VDD pin (3) plus VDDA
• a connection for every capacitor and resistor
• ST's 10 pF Cstray estimate

The SWD header's ~{RESET}, VTref, GND and GNDDetect are mapped too. The VCAP and Cstray values follow the datasheet references in #783. st.com refused automated downloads, so I could not re-read DS10314 myself.

Capacitor and resistor endpoints use the template-local convention that i2c_pullups and led_indicator already use (C8.1 = the eighth capacitor in component-list order). The design notes say so, because apply_template numbers parts from the schematic's next free reference.

Test. builtin_templates_name_symbols_and_pins_that_exist_in_kicad checks every bundled template:

• Each lib_id must resolve through KiCadSymbolSource and resolve_lib_symbol_flattened_node.
• Each from/to endpoint of the form PREFIX[N].PIN must match a pin name, number or alternate of that part's symbol. A symbol that resolves with no pins also fails, so the check cannot pass vacuously.

When no KiCad library is installed the test skips, unless KICAD_CLI is set. In that case it fails, so the E2E job cannot skip it silently. A step in e2e-kicad.yml runs it there.

Architectural fit

This uses the existing symbol resolution that apply_template itself places parts with. No new mechanism or workaround is added.

Exclusions:

• apply_template says it wires the circuit but only places parts; its connection maps omit passives #784 (apply_template does not wire; passives of other templates have no connections) is untouched.
• usb_c_5v_sink's Device:D_TVS with value PRTR5V0U2X is a two-pin symbol for a four-pin part. That is a content question, not a naming one, and is left as is.

Branch and dependencies

Base branch: main @ 9488e5f. No dependencies. Not part of a series. This PR owns all of #783's acceptance criteria.

Compatibility and safety

• Template content changes: ldo_3v3 places AP2112K-3.3 and two capacitors instead of AMS1117-3.3 and four. A caller that hard-coded the old part list for ldo_3v3 sees different parts.
• No schema, tool or file-format change.

Validation

Changed tool behavior

Behavior	Contract and evidence for this change
Accepted inputs and declared defaults	Unchanged.
Invalid/unsupported inputs and structured errors	Unchanged. Before this change, usb_c_5v_sink and stm32_minimal hit the existing "not found in library" error.
Target, data source and prerequisite state	Unchanged; the template data is what changed.
Observed changes and preserved unrelated objects	All six bundled templates apply to the fixture; kicad-cli loads each result.
Failure before/after mutation, including applied work	Unchanged; placement remains all-or-nothing.
Recovery from partial/uncertain results without repeating applied work	Unchanged.

Windows 11, KiCad 10.0.6 installed:

• cargo fmt --all -- --check: clean
• cargo clippy -p konnect-core --all-targets --locked -- -D warnings: clean
• cargo test -p konnect-core --locked --lib: 1657 passed, 23 failed, 21 ignored. The same 23 tests fail on unchanged main @ 9488e5f on this machine (pcb_components, sch_components, schematic_placement_tests, library::symbol_source_tests). The list was compared line by line and is identical, so the failures are environmental. The new test is the one extra pass.
• cargo test -p konnect-core --locked --test integration_test --test conformance_test: 16 passed
• cargo test -p konnect-core --locked --doc: ok
• The whole workspace was not run locally; CI covers it.