v6.60.0

v6.60.0 (2026-05-22)

Bug Fixes

• fix(bom): preserve PCB FPID; prefer PCB Value under PCB-first contract

Two fixes surfaced by the jBOM-vs-Fabrication-Toolkit cross-project
comparison (20 real KiCad projects under /Users/jplocher/Dropbox/KiCad/projects).

Fix #1 -- pcb_reader: preserve canonical FPID

• src/jbom/services/pcb_reader.py:_parse_footprint_node was overwriting
  the canonical FPID (from the (footprint "Lib:Name" ...) opener)
  with whatever the per-footprint (property "Footprint" ...) field
  carried. That property is the schematic-side hint that flowed in via
  'Update PCB from Schematic' and can disagree with the physical
  footprint placed on the board (e.g. PCB FPID 'VendorLib:0805-CAP' vs
  schematic property 'Capacitor_SMD:C_0805_2012Metric'). Under the
  PCB-first contract footprint_name must describe what's physically on
  the board. The schematic hint is now stashed under
  attributes['schematic_footprint'] for DRC-debug visibility.
• Reproduced in LEDStripDriver, cpOD-updated (C0603K vs C0603).

Fix #3 -- synthesize_bom_components_from_pcb: PCB Value wins

• The value resolver previously preferred the schematic value when
  present and fell back to PCB. Reversed: PCB-first contract says the
  per-footprint (property "Value" ...) on the PCB is the canonical
  value because (a) it is what the assembler reads off the actual
  board, and (b) it survives the cpNode-Xiao-orig pattern where the
  schematic carries a uniform default across many distinct refs that
  have been customised on the PCB (e.g. Q1..Q17 all marked 'BSS138' on
  the schematic while the PCB stores per-Q part numbers). Schematic
  Value remains the fallback for the case the PCB has nothing recorded.

Test coverage

• tests/unit/test_pcb_reader.py: new
  test_parse_footprint_node_preserves_canonical_fpid_over_schematic_footprint_property
  pins down Fix #1.
• tests/services/test_bom_workflow.py: renamed
  test_synthesize_uses_pcb_footprint_and_pcb_value (was
  ..._and_schematic_value) and added
  test_synthesize_pcb_value_wins_across_refs_sharing_schematic_value
  -- a direct regression test for the cpNode-Xiao-orig pattern.

Validation

• Full pytest: 1398 passed (was 1396; +2 new tests).
• Full behave: 47 features / 261 scenarios passed; 0 failed.
• jBOM-vs-FT comparison: 15/20 [OK] (up from 14/20); cpNode-Xiao-orig
  fully collapses; cpOD-updated C0603K vs C0603 diff disappears.

Remaining diffs are policy/style, not jBOM bugs:

• Designator case folding (FT uppercases all designators via
  GetReference().upper()).
• LEDStripDriver footprint name length (FT applies its own regex
  normaliser; jBOM now surfaces the canonical PCB FPID, which is the
  right thing).
• cpOD / cpOD-updated DNP refs missing in jBOM by default (FT runs with
  exclude_dnp=False; jBOM defaults to True). Use 'jbom bom --include-dnp'
  to match.

Refs #289.

Co-Authored-By: Oz oz-agent@warp.dev (59de8a2)

• fix(bom): BOM generation is PCB-driven (PCB-first cutover)

BOMWorkflow now follows the per-artifact contract: PCB is the canonical
row set, schematic is enrichment. Built on Stream 1's pcb_project_loader
helpers so BOM and POS share identical resolve/load/enumerate plumbing
(only the artifact_name label differs).

Behaviour changes:

• _generate uses resolve_pcb_input(artifact_name='BOM'); the diagnostic
  trio now reads 'Note: BOM generation requires a PCB file. ... / found
  matching PCB ... / Using PCB ...'. Previously the workflow emitted
  'requires a schematic file' and resolved PCB input back to the sibling
  schematic.
• board.footprints is the canonical row set.
• New helper synthesize_bom_components_from_pcb(board, schematic_components)
  produces schematic-shaped Component records keyed by PCB reference:
  footprint/package/side/mount_type/smd come from the PCB, value comes
  from the schematic when available (falls back to the PCB Value
  attribute), dnp propagates from the schematic, and inventory-biased
  fields are populated later by InventoryOverlayService.
• References present only in the schematic are intentionally invisible
  to BOM/POS - schematic/PCB sync is ERC/DRC territory.
• The post-hoc band-aids enforce_bom_device_footprints and
  enrich_bom_smd_from_project_pcb are no longer invoked from _generate;
  their inputs are now PCB-sourced from the start. The helpers remain
  defined for downstream tooling but are unused by the BOM workflow.

Tests:

• tests/services/test_bom_workflow.py monkeypatches the new
  pcb_project_loader helpers and asserts on the PCB-first diagnostic.
• Four new behavioural tests for synthesize_bom_components_from_pcb
  (schematic-biased value, PCB-biased footprint, schematic-only refs
  invisible, PCB Value fallback, schematic DNP propagation).
• tests/unit/test_cli_verbose.py fixture now includes a .kicad_pcb (BOM
  requires it) and assertions accept PCB-first diagnostic patterns.

1391 focused tests pass.

Refs: #281

Co-Authored-By: Oz oz-agent@warp.dev (df28d4a)

Features

• feat(cli): jbom fab --archive-name parity with plugin

Adds --archive-name TEMPLATE to jbom fab so the CLI produces the same
gerber archive stem as the plugin for the same project. Default
template (_) matches the plugin dialog's editable
Archive field.

Implementation:

• New jbom.services.project_metadata.expand_archive_template(template,
  pcb_file) helper resolves KiCad title-block variables (${TITLE},
  ${REVISION}, ${DATE}/${ISSUE_DATE}, ${CURRENT_DATE}, ${COMPANY})
  against the PCB title block read via the file-based S-expression
  parser. Falls back to the PCB filename stem when the template
  expands to nothing usable; returns '(unknown)' only when no PCB
  exists on disk. Uses a filename-safe normaliser that preserves dots
  so revisions like '1.0' survive (the existing
  normalize_archive_stem stripped non-alphanumeric chars).
• DEFAULT_ARCHIVE_TEMPLATE constant published from project_metadata so
  both CLI and plugin share one source of truth.
• FabricationRequest gains an archive_template field. The workflow's
  _resolve_archive_stem now resolves in this priority:
  1. request.archive_stem (plugin pre-expands and passes the stem)
  2. request.archive_template (CLI passes the template; workflow
     expands once it has the resolved PCB path)
  3. legacy normalize_archive_stem(project_name) fallback
• CLI fab handler reads --archive-name CLI arg, then the saved
  archive_name_template from .jbom/jbom-options.json (plugin and CLI
  share the persisted options file), then DEFAULT_ARCHIVE_TEMPLATE.
• The plugin dialog's existing _expand_archive_template_from_file
  remains unchanged and is functionally equivalent.

Result: 'jbom fab' run against the same project as a plugin click
produces the same archive name (e.g. cpNode-Xiao-68x90_1.0.zip vs the
previous CLI's cpNode-Xiao-68x90.zip).

Tests: 5 new tests in tests/unit/test_archive_template.py cover:
title-block expansion, default template fallback, PCB-stem fallback
when title block is empty, '(unknown)' when PCB missing, and unsafe-
character normalisation.

Refs: #287

Co-Authored-By: Oz oz-agent@warp.dev (c2fbc15)

Refactoring

• refactor(application): factor shared PCB-as-row-set plumbing into pcb_project_loader

New jbom.application.pcb_project_loader module centralizes the
resolve-PCB-input / load-board / collect-project-graph sequence that
BOM and POS each used to open-code.

API surface:

• resolve_pcb_input(input_path, *, artifact_name, options) - wraps
  ProjectFileResolver(prefer_pcb=True, target_file_type='pcb'); emits
  the canonical 'Note: requires a PCB file. / found matching
  PCB. / Using PCB.' diagnostic trio so BOM and POS produce identical
  wording (only the artifact label differs).
• load_board(pcb_path) - thin wrapper around
  DefaultKiCadReaderService().read_pcb_file(...).
• list_hierarchical_schematic_files(project_context) - existence-checked
  enumeration that returns [] on resolver failure.
• load_schematic_components(files, *, options, verbose) - schematic
  enrichment loader; per-file failures become warning diagnostics in
  verbose mode and are swallowed otherwise.
• collect_project_graph(...) - delegates to ProjectComponentCollector
  with board.footprints as the canonical PCB input.
• ResolvedPcbProject - frozen dataclass bundling the resolved input,
  pcb path, project context, and accumulated diagnostics.

POSWorkflow refactored to use the helpers; no behaviour change. _list_fields
and _generate no longer hand-roll the resolver + reader + schematic
enumeration. Existing POS workflow tests monkeypatch the new helper
functions instead of the removed direct ProjectFileResolver /
DefaultKiCadReaderService attributes.

New tests/application/test_pcb_project_loader.py covers resolve_pcb_input
trio behaviour, artifact_name wording, missing-context fatal-error,
list_hierarchical_schematic_files filtering, load_schematic_components
verbose vs silent failure, and collect_project_graph using board
footprints as the canonical input.

Refs: #280

Co-Authored-By: Oz oz-agent@warp.dev (b4fceab)