Skip to content

Releases

v0.29.0

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 10 Nov 00:51

Upgrade notes / behaviour changes

  • Exported .kicad_dru ampacity floors change (Issue #8612). Waived/reinforced nets and oversize minimums are no longer emitted, and designer DRUs with Reviewed clearance rules get none; re-run kicad-cli pcb drc against the shipped DRU. Use --no-ampacity-rules to emit none (previously generated ampacity rules in the managed floors block are then removed on rewrite).

  • kct optimize-traces may now roll back more nets (Issue #8632): nets whose rewrite opens a connection or adds an un-waived creepage pair are reverted, and --layers/--copper default to the board's values instead of 2 / 1.0 oz.

  • A tapeout can newly refuse (Issue #8633): nonzero redundant_copper_* / via_dangling warnings remaining after cleanup now fail the warning review unless acknowledged with kct readiness --ack-trace-quality RULE=REASON.

Fixed

  • The pure-Python fallback search now charges its HV mask setup to its per-net
    budget, and builds those masks in O(N)
    (Issue #8414). On HV boards the
    pairwise-widening masks used to be built before Router._route_impl started
    its clock, so a 10 s fallback call could take ~128 s. The clock now starts at
    the top of the call, and an expired budget returns a deadline stop
    (last_route_deadline_hit) without caching a partial mask. The zero-padded
    disc dilation behind those masks is now an exact separable distance
    transform: byte-identical output, ~0.13 s per mask on a 15.9 M-cell grid at
    any radius, down from 1.2-15 s.

  • Pure-Python router no longer runs on MLX/CuPy grid arrays (Issue #8609). Router now moves a GPU-backed grid to NumPy up front (speedup not measured here), and RoutingGrid._select_backend reports the real backend type for ArrayBackend (it previously logged cpu for an MLX grid). The default kct route path builds grids without a PerformanceConfig, so this is a defensive guard there; the cause of the reported slowdown is not established by this change. cupy-cuda12x carries a sys_platform != 'darwin' marker so uv sync --all-extras resolves on macOS.

  • kct export no longer writes unmeetable or waived ampacity floors into the designer .kicad_dru (Issue #8612). Nets with a waived ampacity/path_ampacity finding or a current-path record get no Ampacity Min Width rule, minimums wider than the board are skipped with a warning, and a designer DRU carrying its own clearance rules is left without ampacity floors. Export now reports whenever it modifies the .kicad_dru. New --no-ampacity-rules flag.

  • --waive no longer fails wholesale on one unmatched key; same-digest re-binds keep the original approval (Issue #8614). kct check --waive and kct detect-mistakes --waive record the keys that match and warn about the rest (exit 1 only when no key matches). Re-waiving a key whose evidence digest is unchanged under a new evN: prefix keeps the original reviewer, date, reason and issue and adds rebound_at. kct check summaries gain orphan_waivers beside stale_waivers.

  • kct place-silk-refs now agrees better with kicad-cli pcb drc and keeps labels with their own part (Issue #8628). NPTH mounting holes are obstacles, the planner adds a measured 0.225 mm slack for KiCad's larger stroke-font text extents (planner-only; kct check is unchanged), and candidates are chosen to be at least 0.4 mm nearer their own courtyard than any other part's. When no clear position reaches that margin the best one is used and the placement is reported as low_association.

  • kct optimize-traces no longer reports a clean result over a newly opened net or creepage violation (Issue #8632). With --drc-aware the written board is re-checked with kicad-cli (zone/via-aware unconnected items per net) and, with --voltage-map, with the kct creepage audit; nets that regress are rolled back and listed under output_verify_rolled_back. creepage_gate is active only when the whole-board audit ran. --layers/--copper now default to the board's own layer count and outer copper weight (stack_source in the JSON report).

Changed

  • Tapeout/readiness cleanup now runs dedupe, prune and optimize-traces --drc-aware, and gates on trace quality (Issue #8633). kct readiness --generate records per-step counts in README.txt and the manifest, and optimize-traces reads layer count and copper weight from the board/project when --layers/--copper are not given.

Performance

  • Negotiated routing stops re-probing known-failing signal nets (Issue #8486). The no-sealed-pad
    fallback key is now the set of foreign (net, layer) pairs near the failed net instead of exact
    geometry, so the rescue memo survives a re-route pass (re-probed every 3 iterations), and the
    per-net search back-off also applies to nets that fail before reaching 90 % of their cap.

v0.28.1

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 08 Nov 19:23

Fixed

  • Post-negotiation sweep stops at its bound inside pad-escape and sub-grid
    retries
    (Issue #8414). A stranded wide power net whose edges fail on pin
    access no longer runs pad-escape generation, the sub-grid retry or the
    via/trace conflict resolver on an already-spent per-net or whole-sweep
    budget: each step checks the remaining budget before it starts, escape
    generation stops between pads, nested retries get the remainder, and the
    rescue DRC pre-check stops at the sweep deadline (rolling the rescue back).
    The sweep's timing line now says what it enforces, and a net that runs more
    than 1 s past its per-net bound is named with where the time went.
    Unbudgeted / deterministic-budget runs are unchanged.

v0.28.0

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 07 Nov 19:05

Upgrade notes / behaviour changes

  • kct export can newly exit non-zero (Issue #8203). Environments without the report extra must install kicad-tools[report] or pass --no-report.

  • kct optimize-traces now loads the board for collision checking (Issue #8214). A board the router loader cannot handle (for example one with no Edge.Cuts outline) now exits 1 with a clear error instead of being optimized unchecked; the loader's cost is also paid on --dry-run.

  • New redundant_copper warnings; readiness cleanup step (Issue #8218). Boards may newly show redundant_copper_* warnings (never errors). kct readiness --generate now mutates the candidate PCB by removing exact-duplicate copper; pass --skip-cleanup to opt out (the skip is recorded in README.txt).

  • kct creepage --waive-same-footprint can newly fail (Issue #8232). Net pairs whose minimum is same_footprint now carry an extra board row; if that board-level approach is below the requirement the gate fails despite the waiver. JSON/table output may contain two rows for one net pair.

  • --waive-attach-zone can now fail boards that passed (Issue #8292). Foreign copper outside a part's attach zone that is closer than the requirement to the part's in-zone pad now gates the pair. Move that copper or review the new residual_witness_mm.

  • kct optimize-traces --net is now an exact, case-sensitive match (Issue #8299). It previously matched any net whose name contained the text, case-insensitively. Scripts relying on that should switch to --net-pattern '*GND*' (--net and --net-pattern are mutually exclusive).

  • kct route-link commit mode now needs the pinned kicad-cli (Issue #8314). Without --probe it refuses (exit 2, nothing written) when the pinned KiCad toolchain is unavailable; set KCT_KICAD_DOCKER=1 to use the pinned image. --probe is unchanged.

  • kct optimize-traces --drc-aware can now exit 2 (Issue #8322) when kicad-cli reports more clearance errors on the written board than on the input. The board is still written.

  • pcb prune --current-paths default changed: --wide is now net-relative when omitted (was an absolute 0.5 mm). Pass --wide 0.5 to restore the old behaviour.

  • Action needed if you run pcb prune, optimize-traces --drc-aware or stitch --drc on a board with no .kicad_pro/.kicad_dru beside it (Issue #8369). They now fail instead of silently checking against KiCad default rules. Pass --project PATH.kicad_pro, or --allow-default-rules to keep the old behavior.

  • pour_slot is a new kct check warning category and never fails strict board gates (Issue #8435). It is listed in NON_GATING_RULE_PREFIXES alongside redundant_copper, so board gates that fail on any warning (board 04, board 05) count it as an advisory. Run kct check --skip pour_slot to silence it; board 04's two current hits (USER_LED/NRST, GND/SWDIO on F.Cu, 0.31–0.33 mm gaps) are ordinary track spacing, not a missing plane.

  • Pour-slot packing is now opt-in (Issue #8457). Add --pour-slots-pack to kct optimize-traces --pour-slots (or pack=True to cleanup_pour_slots) to keep the old pack fallback.

  • Recorded waivers go stale once (Issue #8517). EVIDENCE_HASH_VERSION is
    now ev3 (mistake hashes ev3.m1), so every ev2 waiver reports
    STALE waiver (outdated evidence version), even on boards whose origin is
    (0, 0). Re-review and re-record with kct check --waive.

  • Ampacity rules in the exported .kicad_dru change shape (Issue #8527). Per-net targets now emit A.NetName == '...' rules rather than one A.NetClass rule at the class maximum, and widths follow the stackup copper weight. Re-run kicad-cli pcb drc against the shipped DRU after re-exporting.

  • Regenerate any dual-row header footprint previously made with kct footprint generate pin-header --rows 2: its pad numbering was column-major and is now row-major (stock KiCad numbering). (Issue #8529)

Fixed

  • kct export fails when the report extra is missing (Issue #8203). A missing report dependency used to log one warning and exit 0 with no report.md/images/; it now records an error naming the module that failed to import plus the kicad-tools[report] install hint, and exits non-zero.

  • kct report generate now loads the committed net-class-map sidecar (Issue #8204). Sidecar-gated DRC rules (diffpair_length_skew, diffpair_routing_continuity, diffpair_uncoupled_length, match_group_length_skew) no longer run inactive in the report's DRC section, matching kct export and kct check.

  • kct optimize-traces --drc-aware reports attempted vs kept nets and why nets rolled back (Issue #8207). Summary now reads "Nets attempted: N (kept: K, rolled back: R)"; -v and --format json list per-net new_violations by rule id, the failsafe warning names the rules above baseline, and a fully reverted run writes the input board byte-for-byte.

  • kct parts availability now matches kct export and accounts for every BOM line (Issue #8208). LCSC-less lines are auto-matched like export (opt out with --no-auto-lcsc; matches are labelled lcsc_source in JSON and [auto-matched] in the table), and the table/summary now report and list stock-unverified, lookup-unavailable and discontinued parts so the counts sum to the total.

  • kct optimize-traces now works on KiCad-10 name-only boards (Issue #8209).
    On a board with no (net N "name") header table, replace_segments removed
    none of the original segments and appended the optimized copper as (net 0)
    no-net segments, which created shorts. Under --drc-aware, every net was
    rolled back. Segments are now matched by resolved net name and written in the
    board's own (net "NAME") dialect.

  • kct optimize-traces collision-checks every rewrite on finished boards (Issue #8214). The CLI now builds a collision checker from the board's own copper and clearances, so unsafe chamfers/zigzag/staircase rewrites are skipped individually instead of --drc-aware rolling back whole nets. Each net's accepted rewrite is synced into the checker's grid before the next net is checked, and KiCad-10 name-only boards are keyed to the grid's net ids. If the checker cannot be built the command fails instead of falling back to unchecked rewrites.

  • Via minimisation no longer strands pad/pour connections (Issue #8215). kct route now gives the via optimiser pad positions and same-net zone layers, and via-pair elimination is connectivity-checked. kct optimize-traces never runs via minimisation.

  • kct optimize-traces now shortens detours between fixed points (Issue #8217). The shorten_detour_pass that kct route already ran now also runs from optimize-traces (incl. --net), using the loaded board's pads and vias as anchors and its copper as the obstacle model. Length-matched nets are skipped.

  • kct optimize-traces off-angle warnings name the net and no longer repeat (Issue #8219). OffAngleSegmentWarning now reports net GND instead of net 0 on name-only boards, and intermediate DRC-trial serializations no longer re-warn for each segment.

  • deploy-site.sh leaves a clean git tree (Issue #8220). Tracked renders
    rewritten by kct render are restored after the site build (published bytes
    unchanged), .wrangler/ is gitignored, and a warn-level check reports any
    remaining uncommitted paths so Pages commit provenance stays accurate.

  • kct creepage no longer hides board-level approaches behind a same_footprint row (Issue #8232). When a net pair's minimum gap is component-internal, the census now also emits a second board row with the smallest residual board-level creepage (sharing footprint's pads masked out), so --waive-same-footprint waives only the intra-footprint row.

  • PCB.save no longer adds a numbered net table to name-only boards (Issue #8239). Loading and saving a KiCad-10 board with no (net N "name") header now leaves it header-less; kct pcb prune inherits the fix.

  • kct pcb prune --no-drc --unverified no longer leaves a dangling end or cuts an arc-fed pad (Issue #8249). Removal now rejects groups that strand a neighbouring track end resting on a removed body, and copper arcs protect any segment whose body reaches them.

  • kct optimize-traces no longer breaks connectivity on finished boards (Issue #8250). The 45° pass leaves corners that a via, pad or T-junction branch attaches to; keepout rule areas and NPTH holes are now honoured by the collision check; a net whose rewrite would split previously-joined pads is discarded and reported (nets_connectivity_discarded); and --drc-aware no longer rolls nets back on clearance_segment_zone, which is measured against stale stored zone fills. Pad-to-zone connections are not covered by the connectivity gate.

  • kct route --timeout report no longer names a moved checkpoint (Issue #8258). With a relative -o, <stem>.timeout.json checkpoint now points at the quarantined _timeout_unverified_ file, and --export-failed-nets records in the report when it was skipped on timeout. The report names the search-failures sidecar only if one was written before the deadline, and gives a reason otherwise.

  • kct pcb prune removes a redundant branch whose free end rests on a via by overlap only (Issue #8259). An anchorless track end whose capsule overlaps a same-net via/pad is now overlap-held (soft protection) instead of a hard free end; the overlap is still never counted as a join. When the branch goes, the inner-layer end that shared the overlap is moved onto the via/pad centre, or, if that would tilt a 45-degree track, gets a short same-layer stub from the centre (new stubs_added stat).

  • kct check --output JSON now carries summary.attach_zone_exemption (Issue #8262). The file report previously omitted the attach-zone exemption state (and _reason / `_unexem...

Read more

v0.27.0

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 31 Oct 19:08

Upgrade notes / behaviour changes

  • Bare route sidecars are no longer written (Issue #8156). Consumers that
    expected fab_profile.json / net_class_map.json after kct route must read
    <out-stem>.fab_profile.json / <out-stem>.net_class_map.json. kct check
    still discovers existing bare files.

Fixed

  • kct audit / export readiness report honour the waivers sidecar (Issue #8130). Waived DRC findings no longer count as blocking (matching kct check); they are listed separately as waived with rule, count and reason. Stale evidence-bound waivers still leave the finding blocking.

  • Suppressed attach-zone exemption is now visible (Issue #8131). When a designer clearance rule in .kicad_dru keeps kct from writing its managed attach-zone block, kct route and kct check print a warning (check includes the count of unexempted findings), and kct check --format json adds summary.attach_zone_exemption (active / suppressed / not_applicable) with attach_zone_exemption_reason. The suppression rule itself is unchanged.

  • LCSC discrete-footprint guard no longer FAILs DIP bridge rectifiers (Issue #8132).
    A DIP-N part on a Diode_* footprint whose name carries the same DIP-N token
    (e.g. Diode_THT:Diode_Bridge_DIP-4_W7.62mm_P5.08mm) now passes; a different pin
    count or a footprint with no DIP token still FAILs.

  • kct drc rejects JSON that is not a DRC report (Issue #8133). A mis-pointed or non-report JSON file (e.g. {"anything": 1}, a list) previously loaded as zero findings and passed with --allow-incomplete; it now exits 1 with Error loading report:.

  • kct pcb move-via --check now catches hole violations (Issue #8153). Besides copper clearance, --check evaluates the moved via's hole_clearance (drill edge to foreign copper on every spanned layer) and hole_to_hole against the board's .kicad_pro/.kicad_dru, refusing moves kicad-cli would reject. The JSON check block lists rules_checked.

  • bom_lcsc_stock preflight no longer reports OK when it checked nothing (Issue #8154).
    LCSC numbers merged from an existing bom_<fab>.csv are now evaluated (also for
    bom_lcsc_values), and when assigned numbers exist but none can be found in the parts
    cache or offline catalog the check warns (fails under --strict-preflight).

  • Router via-blocked diagnostics name the local blocker (Issue #8155). A C++
    decline "blocked by stored-via geometry (blocking net X)" now reports the net
    nearest the start/goal pad instead of whichever via candidate was expanded
    last, and pin-access failures caused by via-site exhaustion print a "PIN
    ESCAPE BOXED IN" suggestion rather than the generic --grid advice.

  • RV* varistors no longer fail the tapeout LCSC value gate as resistors (Issue #8157). parse_component_value now classifies RV references as OTHER instead of RESISTOR, so a MOV with a 220pF@1kHz description is "cannot validate" rather than a cross-type FAIL.

  • /kct:manufacturing-readiness and /kct:tapeout Gate 1 no longer refuse on a stale manifest (Issue #8158). Re-running sign-off into an existing bundle directory now clears the previous manifest.json before kct check, so the Manifest sub-check reads NOT RUN instead of FAILED.

  • silk_overlap no longer flags a footprint's lead line meeting its own body rectangle (Issue #8159). A same-footprint fp_line ending on an fp_rect outline is one drawing, like the line-vs-line joints from #8090; crossings, edge overlaps and cross-footprint pairs still flag.

Added

  • kct pcb move-via nudges a free via and drags its copper (Issue #8135). Select a
    via by --at X Y (with --tolerance), --net or --uuid, move it with --to/--by;
    coincident trace endpoints follow. Supports --dry-run, --format json, and --check
    to refuse a move that adds a clearance violation.

Changed

  • Vendored check_routed_drc.py no longer carries board04 logic (Issue #8134).
    The paid-drill dispatch moved to the kicad-tools-only
    scripts/ci/check_routed_drc_internal.py; re-running install-kct.sh
    replaces the consumer copy with the generic gate.

  • kct route --timeout reserves more time for finishing (Issue #8148). The
    post-route reserve is now 27% of --timeout (cap 30%, floor 60 s), e.g.
    162 s at --timeout 600, so zone refill and DRC fit after routing; the
    chosen reserve is logged.

  • kct route sidecars are keyed to the output stem (Issue #8156). It now
    writes <stem>.fab_profile.json and <stem>.net_class_map.json instead of the
    bare fab_profile.json / net_class_map.json, so boards sharing a directory no
    longer collide. A non-zero exit lists any *_unverified_* sets it left behind.

v0.26.0

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 30 Oct 20:28

Upgrade notes / behaviour changes

  • install-kct.sh without --tag/--rev/--path now switches a consumer
    to PyPI
    (Issue #8077). It previously pinned a git tag
    ([tool.uv.sources] kicad-tools = { git = ..., tag = "v<version>" }); a
    re-run now removes that entry and writes kicad-tools[...]>=<version>.
    Run uv lock afterwards and commit uv.lock. To keep a git pin, pass
    --tag v<version> (or --rev <sha>) explicitly. Requires Python >= 3.11 (or python3 with tomli) on PATH, or one uv python find can locate.

  • Changelog Fragment Check can newly fail on commit trailers (Issue #8099).
    A PR whose commits say Part of #<epic> must cite that epic in a fragment,
    e.g. (Issue #7423, part of #5798), or carry the changelog:skip label.

  • kct export can newly warn or fail on out-of-stock LCSC parts (Issue #8121).
    A part whose cached/offline stock is below placements x 5 boards (e.g. a
    stock=0 C4223175) now WARNs, and FAILs under --strict-preflight or when
    the stock was observed live. Pass --boards N to match your real order size.

Fixed

  • install-kct.sh PyPI pin hardening (Issue #8097). The pin helper now
    keeps a CRLF (or mixed-ending) pyproject.toml line-for-line instead of
    rewriting it to LF, and decodes TOML escapes such as "kicad\u002Dtools>=0.20"
    so it rewrites that requirement rather than appending a duplicate.
    --dry-run and already-current repos need only Python >= 3.9; Python >= 3.11
    (or tomli) is required just for an actual rewrite. The vendored
    kct-upgrade.yml checks out with persist-credentials: false, so the
    installer and uv lock never see a token; the PR step pushes with its own
    step-scoped GH_TOKEN.

  • Release changelog gap gate no longer flags issues already documented in the window's starting release (Issue #8111). When changelog_gap_report.py runs with --since vX.Y.Z, issues cited in the ## [X.Y.Z] section now count as documented, matching the per-PR check.

  • LCSC guard no longer passes an IC on a discrete footprint (Issue #8113). parse_ic_package now recognises BGA/LGA/CSP packages so the pad-count check runs, and a new rule flags an IC package or IC category on a Diode_*/Resistor_*/Capacitor_*/Inductor_*/LED_* footprint even without PCB pads. /kct:tapeout --assembly Gate 4 now reads the manifest's bom_lcsc_values result and refuses on FAIL, a missing result, or unverified lines.

  • kct-upgrade workflow keeps its write token away from installer code (Issue #8115). The
    template now runs the installer and uv lock in a read-only prepare job and
    hands a patch to a separate publish job, the only one holding a token. The PyPI pin helper
    also rewrites multi-line ("""...""" / '''...''') requirements in place instead of
    appending a duplicate, and rejects the TOML 1.1-only \e / \xHH escapes like tomllib does.

  • Anonymous JLC parts lookups no longer 403 after the first request of a session (Issue #8118). The parts endpoint sets an XSRF-TOKEN cookie on its first response and then refuses (403) every POST that carries the cookie without echoing it as an X-XSRF-TOKEN header. LCSCClient kept the cookie but never sent the header, so kct parts availability, lookup_many and BOM enrichment resolved only one part live, tripped the 403 circuit breaker, and reported the rest as unavailable or offline-only (it looked like a geo-block). The header is now sent, and a 403 while a token is held drops the cookies and retries once tokenless before tripping the breaker (the token expires after 1800 s).

  • LCSC discrete-footprint IC guard no longer fails diodes or caps that mention "logic" or "memory" (Issue #8124). The IC-category clause now matches whole LCSC category phrases (Logic Gates, Non-Volatile Memory, Microcontrollers, ...) on word boundaries instead of bare substrings, so a description like "ESD protection diode for logic lines" no longer FAILs bom_lcsc_values and blocks /kct:tapeout --assembly. The IC-package check is unchanged.

  • LCSC IC-on-discrete-footprint guard recognises pin-count-first BGA and LFCSP packages (Issue #8127). 48-TFBGA (6x8) now parses as BGA-48 and LFCSP-32-WQ(5x5) as a QFN-group package, so such parts assigned to a discrete footprint are flagged.

Added

  • kct export checks LCSC stock against the order (Issue #8121). New
    bom_lcsc_stock preflight check compares each assigned part's recorded
    stock with placements-per-board x --boards N (new flag, default 5, JLCPCB's
    minimum). A shortfall on live-verified stock FAILs; on stale/offline/unknown
    stock it WARNs (FAIL under --strict-preflight). Also surfaced by
    validate_assembly_bom and kct readiness.

Changed

  • install-kct.sh installs kicad-tools from PyPI by default (Issue #8077).
    A bare install-kct.sh <target> (or --pypi) now writes
    kicad-tools[<extras>]>=<version>, keeps the target's existing extras,
    removes any [tool.uv.sources] kicad-tools entry and records
    source_mode: "pypi"; uv.lock is the exact pin. --tag, --rev and
    --path are unchanged and documented as escape hatches. Only kicad-tools
    requirements inside dependency arrays are edited, never keywords,
    descriptions or comments. kct doctor treats a >=X floor as satisfied by
    any installed version at or above it. It reports drift when the installed
    package is below the floor (fix: uv sync), or when the vendored skills and
    gates (install-metadata.json) record a different version than the package
    (fix: re-run the installer). New opt-in --upgrade-workflow vendors
    .github/workflows/kct-upgrade.yml, a weekly job that opens a PR moving the
    consumer to the latest PyPI release (never a downgrade) and lists that
    release's upgrade notes.

  • The per-PR Changelog Fragment Check now reads the PR's commit trailers
    (Issue #8099). A user-visible commit whose Part of #N (or Closes #N /
    Fixes #N) issue no fragment cites now fails the check at PR time, with the
    one-line fix in the message ("add 'part of #N' to "), instead of
    passing and then turning the release gate red after merge. Internal-title
    PRs, internal commits and INTERNAL_ISSUES entries are unaffected.

v0.25.1

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 29 Oct 21:10

Fixed

  • Pour completion cut by --timeout is reported as its own outcome (Issue #8078). When the KiCad-oracle pour stage stops on the deadline with links still open, kct route now logs the time-left / DRC-cost / attempt-estimate that made the gate refuse and says the closer was cut by --timeout instead of presenting the links as plain unconnected GND/VCC nets.

  • Bench rows label a pour-completion deadline cut (Issue #8078). expand_route_experiment.py now records pour_cut_by_timeout and the route table's Timeout cell reads yes (pour completion cut), so open GND/VCC links caused by --timeout are distinguishable from a closer failure. The manufacturable verdict is unchanged.

  • kct pcb reinforce honors the board's own clearance rules (Issue #8085).
    Anchors are now checked against the net-class clearance floors from the
    .kicad_pro (as kct check reads them), and with the new --voltage-map
    (plus --creepage-standard, --pollution-degree, --material-group,
    --hv-threshold) against the pairwise IEC requirement kct route uses. The
    hole-to-hole check now includes same-net PTH pads and vias, so anchors no
    longer land on pad drills. Refusals report the governing rule
    (hole_to_hole / dru_class_floor / voltage_map_pairwise) with required
    vs actual mm in text and JSON.

  • silk_overlap no longer flags T-joints within one footprint (Issue #8090).
    A line ending mid-span on another fp_line of the same footprint is treated
    like a corner join; duplicates, collinear overlaps, X-crossings and
    cross-footprint pairs still flag.

  • kct pcb current-paths-audit no longer reports a correct Kelvin layout as ambiguous (Issue #8092). A thin sense tap off an endpoint pad (stub -> via -> sense-IC pad, acyclic and narrower than the force copper) is now set aside from the endpoint-fanout judgement and listed in PathResolution.sense_taps instead of being treated as a force leg; rejoining, equally wide or dangling taps and real via arrays stay ambiguous.

v0.25.0

Choose a tag to compare

@loom-fleet-dispatch loom-fleet-dispatch released this 29 Oct 05:01

Upgrade notes / behaviour changes

  • kct route holds more via keep-outs by default (Issue #4410). Before
    routing, a +5V/GND-style decoupling cap whose pads are all on served pour
    nets now gets a reserved via site. Signal copper may therefore take a
    different path than before on boards with such caps. If the larger
    keep-out costs a signal net on your board, pass
    --no-reserve-pour-only-passives to get the previous reservation back.

  • DifferentialPairRules.for_type is removed (Issue #6448). It has
    emitted a DeprecationWarning since Issue #6443 and was a shim over the
    protocol defaults table. Call DifferentialPairRules.for_pair_type(pair_type)
    instead; it returns the same rules. To pick a protocol row directly, use
    DifferentialPairRules.from_protocol(...).

  • --differential-pairs on the grid engine is deprecated (Issue #6448).
    kct route --differential-pairs with the default --route-engine grid still
    routes pairs through the coupled pre-pass, now with a one-time stderr warning;
    the library entry point route_diffpair_prepass (on Autorouter and
    DiffPairRouter) emits a DeprecationWarning. Migrate to
    --route-engine lattice --strategy basic --differential-pairs (as board 07
    now does). The grid pre-pass is removed one release later.

  • A pair near its uncoupled-length limit can newly fail kct check (Issue #6554). kct's uncoupled figure now follows KiCad's primitive pairing. That figure is larger than the old projection-overlap figure on pairs with stubs at pads or with copper between the traces: board 03 USB_D reads 10.087 mm, up from 7.212 mm. Pairs within a few millimetres of their limit that passed before may now fail, as they already did in kicad-cli. No committed board is affected and no limit changed. kct is still not exactly KiCad on all router copper: on board 06's regression fixture it reads up to 0.21 mm low (Issue #7565). Re-check pairs within 0.3 mm of their limit in kicad-cli.

  • Diff-pair skew limits follow the Q16 per-class table (Issue #7521).
    Pairs whose class declares pair_protocol and no skew_tolerance_mm now
    resolve: USB2_HS 1.0 mm (was 1.27), USB3_SS / PCIE_G1 / PCIE_G2_PLUS /
    HDMI_TMDS / new SATA 0.127 mm (was 0.1, 0.127, generic 0.5, 0.5), MIPI_DPHY
    0.15 mm (was 0.038), and new DDR_DQS_CK 0.15 mm. These rows are now cited,
    so a breach is an error rather than a warning. A pair that passed
    before can newly fail diffpair_length_skew; set skew_tolerance_mm on the
    class to state a different number.

  • Skew bumps are smaller, nearer their source, and plain copper (Issue
    #7521). Pairs with a large skew on a short or distant run may now stay above
    their diffpair_length_skew limit after --length-match-diffpairs, because
    the tuner no longer adds a bump pushing out more than 2x the pair gap (an
    interim limit; issue #7649 tracks lowering it to 1x), or one more than
    15 mm from a pad or via. Bump shape changed too: the tuner's output no
    longer contains (generated ... tuning_pattern) objects. Bumps are
    45-degree-chamfered plain copper, so pcbnew has no tuning pattern to
    regenerate or edit on them.
    Fix such a pair in the route (equalise the legs near the source) or raise
    the class's skew_tolerance_mm.

  • A pair near its uncoupled-length limit can newly fail kct check (Issue #7565). kct's figure is now larger on pairs whose copper joins a trace mid-run or ends a stub on a pad edge, as kicad-cli's always was. A pair that passed by up to ~0.2 mm may now report the same overrun kicad-cli does.

  • HierarchyNode.path and kct sch hierarchy "file" values are now
    absolute
    (Issue #7614). Scripts that compared them with relative paths
    (for example "sub.kicad_sch" or "project/sub.kicad_sch") should compare
    resolved paths or Path(...).name instead. kct sch pin-map --sheet now
    matches the sheet name or the file path relative to the root schematic's
    folder, so a filter no longer matches a folder above the project. A
    sub-sheet file that is a symlink now has its own sub-sheets resolved next
    to the link's target, not next to the link.

  • kct bom row counts drop by the number of virtual symbols (Issue
    #7615). The table, --format csv and --format json outputs no longer
    list power symbols, PWR_FLAGs or in_bom no symbols, so scripts that
    counted rows or total will see smaller numbers. --format jlcpcb now
    applies --exclude and --include-dnp like the other formats. Code
    calling BOMItem.is_power_symbol / BOM.filter() now also treats
    (power)-flagged library symbols and every #-prefixed reference (not
    only #PWR) as virtual.

  • BOM quantities change for multi-unit parts and repeated sheets (Issue #7616). Re-check
    any stored or diffed BOMs: a multi-unit part (e.g. a dual op-amp) on any sheet is now one
    row instead of one row per unit, and each placement of a repeated sheet adds its own rows.
    extract_bom(..., hierarchical=False) still emits one row per unit.

  • Hierarchy report counts include nested sheets under every placement (Issue #7616). Read-only
    reports (kct sch hierarchy, summary, list-labels, and similar) now count a sheet nested
    inside a repeated sheet once per placement, as they already did for directly repeated sheets.

  • kct render reports partial for boards with bare footprints (Issue #7682). A board where any non-exempt footprint has no usable 3D model now gets model_check.status: "partial" with one reason per footprint (model_check.uncovered_models) instead of passed with 0 models checked; the exit code is unchanged. The new kct check model_coverage category is opt-in (kct check --3d-models or --only model_coverage) because its result depends on the host's KiCad 3D library: plain kct check, --strict exit codes, board gates, kct readiness evidence and run_drc are unchanged. It becomes default-on, so --strict can then fail on a footprint with no model, once a pinned model set lands (#7683/#7685).

  • kct check now includes model_coverage in its default set (Issue #7685). A footprint with no 3D model ref, or a placeholder or hidden-only body, is an advisory warning, so --strict now fails on it; a host without a KiCad 3D library prints one NOTE: on stderr instead of a finding, so serialized reports and committed evidence are byte-identical with or without a library. --3d-models is kept as a no-op. Re-record committed kct check evidence for boards you maintain.

  • LCSC 3D-model fetch is capped and announces its terms (Issue #7686). --fetch-lcsc / KCT_LCSC_FETCH=1 now prints a one-time terms notice per cache directory, stops after 25 fetches per run (raise with KCT_LCSC_MAX_FETCHES), waits 1 s between requests (KCT_LCSC_FETCH_INTERVAL) and refuses to write a cache inside a git work tree. add-3d-models also derives LCSC C-numbers from the BOM, so cached bodies can now resolve for footprints that had no sidecar entry. The new tests/test_lcsc_no_commit_guard.py can newly fail if an EasyEDA/LCSC STEP body is tracked under boards/ or src/: remove it from the repo (LCSC models are never committed or shipped).

  • kct render and scripts/deploy-site.sh can newly fail (Issue #7689). Any non-exempt footprint without a resolvable 3D body (including footprints with no (model …) at all) now exits non-zero, and the deploy script aborts instead of falling back to 2D. Run kct models sync, fix the board, or pass --allow-incomplete-models to kct render for exploratory output.

  • Board generators can now exit 1 when kct export fails (Issue #7740). A recipe run in an environment without the parts extra previously exited 0 with a missing bundle; it now reports Overall: FAIL. Install the parts extra in CI jobs that run generators.

  • model_coverage coverage may now read unknown (Issue #7743). JSON
    consumers that treated every coverage entry as checked will see
    unknown:host_library_refs (...) on boards that use KiCad library 3D models.
    It is non-blocking: sign-off and exit codes are unchanged. Use
    kct pcb model-audit for the per-footprint, per-host verdict.

  • kct route may now run an extra completion pass (Issue #7831). A grid route that ends with open signal nets spends extra time on a --complete pass and can exit 0 where it used to exit 2. Pass --no-chain-complete to restore the previous behavior.

  • Diff-pair skew can newly warn, or fail, on layer-asymmetric pairs
    (Issue #7892). kct check now compares propagation delay, so a
    length-matched pair with one leg on an inner layer (about 23 % slower than
    F.Cu microstrip on a JLCPCB 4-layer board) can exceed its budget. On the 06rf
    fixture USB3_TX1 is 0.015 mm out in length but 28.8 ps out in time, against
    a 4.0 ps budget. A pair with a declared pair_protocol is now judged against
    the ps budget instead of the protocol table's millimetres. At the protocol's
    maximum rate the result is a warning: board 03's USB 2.0 FS pairs move
    from a 2.5 mm error limit to a 115 ps warning budget, and 06rf USB2_D from
    a 1.0 mm error to a 14.4 ps warning. Set data_rate_gbps on the net class to
    make the budget an error at the rate the board actually runs. A class's
    own skew_tolerance_mm and the generic 0.5 mm are still checked as length
    limits (the class fields until #7894). The router and tuners still tune to
    millimetres until #7893.

  • Pairs with a derived budget are now tuned to it, and the .kicad_dru skew
    rule no longer keeps the protocol table's millimetres as a floor
    (Issue
    #7893). A pair whose only limit used to be a warning-level table value
    (LVDS) is now tuned to its ps budget. A pair with no class cap is judged
    against its budget instead of the table value (board 06 USB2_D: 14.4 ps,
    about 2.1 mm at the slowest layer, instead of 1.0 mm), so its exported
    kicad-cli rule l...

Read more

v0.24.0

Choose a tag to compare

@rjwalters rjwalters released this 20 Oct 15:33

Upgrade notes / behaviour changes

  • --edge-clearance 0 no longer permits copper across Edge.Cuts
    (Issue #6043). edge_clearance=0 (and an unresolved None) used to turn the
    board-edge keepout off. It now means "no margin beyond not touching the
    cut": the router keeps every trace and via off the outline and interior
    cutouts by a small physical floor (half the widest trace or via plus half a
    grid cell). There is no opt-out. Pads that sit on or very near the outline
    (edge connectors, castellations) now get a thin keepout ring when no clearance
    is configured; set a positive --edge-clearance if you need a specific value.

  • Crashed kicad-cli DRC runs now fail gates that used to pass (Issue #6724).
    These results used to read as clean and now fail. A kct route whose
    post-route kicad-cli DRC crashed or was killed fails the DRC gate even
    without --strict-drc. A route whose first pour-oracle DRC did not
    complete exits 10 (unverified) instead of 0. That includes a first
    oracle DRC that times out (the default per-call budget is 120 s), not only
    a crash. The recipe pipeline gate fails the DRC leg even with
    require_drc=False, which still excuses only a missing kicad-cli. If you
    start seeing these, read the reason: a signal kill (SIGKILL (out of memory?)) usually means too many Docker DRC containers at once, so run
    DRC one board at a time; an "unsupported option" (kicad-cli printed
    Unknown argument: <opt>) means your kicad-cli is too old for an option
    kct passed; exit 3 ("could not load the board") usually means the board
    was saved by a newer KiCad than your kicad-cli, so upgrade KiCad. Any
    other exit status is reported as-is, with kicad-cli's stderr.

  • KiCad 8/9: geometric DRC now actually runs, without zone refill
    (Issue #6724). Before, kct passed KiCad 10's --refill-zones to every
    kicad-cli; KiCad 8/9 rejected it and the rejected run read as a clean
    DRC. Now KiCad 8/9 run the check without --refill-zones, so kct route,
    kct audit, the recipe gate and benchmark metrics may report geometric
    violations that were previously hidden. KiCad 10+ behaves as before
    (zones are refilled before checking).

  • diffpair_length_skew and match_group_length_skew numbers shift (Issue #6725). Via length now follows the board stackup and the layers each via connects on, and pad/via entry geometry is trimmed, so reported skews change by a few hundredths of a millimetre per via (more where a via declared F.Cu/B.Cu only carries tracks on an inner layer). Pairs and groups can newly fail or newly pass; boards routed with an earlier kct may need kct route re-run to be retuned against the new measure.

  • A via connected on only one layer no longer adds length in kct check (Issue #6725). kicad-cli gives a via no height when the tracks ending in it (and any pad overlapping it: a surface pad's own layers, a through-hole pad's footprint side) use only one copper layer. diffpair_length_skew and match_group_length_skew now do the same on the written board, and no longer charge the via's declared span. A pair or group whose skew came from such a stub via can newly pass, or newly fail. The router's tuners still charge the declared span, because one routed fragment of a net may not show the via's other layer.

  • Large boards routed with a fine --grid on the Python backend now use
    that grid
    (Issue #6741). A run that was silently coarsened to ~500k cells
    now allocates the full requested grid (e.g. 6.05M cells for 195 x 155 mm at
    0.1 mm) and is correspondingly slower and larger in memory. Build the C++
    backend (kct build-native), or pass a coarser --grid that still satisfies
    your clearance, if the run is too slow.

  • New kicad-cli DRC failure reasons (Issue #6775). Scripts that match
    GeometricDRCResult.reason should know that exit 3 (board failed to load)
    is now kicad_cli_load_failed. Every other non-zero, non-signal exit is
    now kicad_cli_exit_error. Both used to be kicad_cli_crash, which now
    means only a signal kill or an exception. All three are still failures
    (GeometricDRCResult.failed), so gates behave as before.

  • kct route holds coupled P/N traces to the pair gap, not the class
    clearance
    (Issue #6810). For a coupled class whose pair gap is larger
    than its clearance, the pre-save check can now fail on P/N traces closer
    than the pair gap -- the copper kct check (diffpair_clearance_intra) and
    kicad-cli (diff_pair_gap) already reject.

  • kct route --differential-pairs can spend longer on pairs it cannot
    couple
    (Issue #6835). Without --diffpair-per-pair-timeout, the coupled
    search's only bound is its iteration budget, which is now 1,400,000 per pair
    instead of 40,000 (PCIE_TX, PCIE_RX and LVDS<n> pairs keep 40,000). A pair that
    fails both phases can take on the order of 15-20 s instead of about 0.5 s.
    Set KCT_COUPLED_FLAGOFF_MAX_ITERS=40000 to restore the old cap for every
    pair, or pass --diffpair-per-pair-timeout to bound each pair by wall clock.

  • kct route --grid <value> now refuses grids past a per-backend cell
    ceiling
    (Issue #6875). Since #6741 an explicit grid is never coarsened,
    so a very fine grid on a large board could allocate until the machine ran
    out of memory. An explicit grid needing more than 20,000,000 cells on the
    Python backend, or 100,000,000 on the C++ backend (counted at the largest
    layer count the run would try), now exits 1 before routing with an error
    naming the cell count and suggesting kct build-native, the finest
    --grid that fits, or --grid auto. Pass --force to route at the
    requested pitch anyway. --grid auto is unaffected, and no committed
    board recipe comes near either limit (the largest, board 09's
    --grid 0.05 investigation, is 15.6M cells at 6 layers).

  • MCP clients must not send undeclared arguments (Issue #6914). tools/call
    requests carrying keys that are not in a tool's inputSchema now return a
    validation error; remove the extra keys.

  • kct pcb import-ses can now fail on an edited board (Issue #6915). If
    the board changed after export-dsn, re-export and re-route, or pass
    --allow-stale. Boards with no sidecar still import, with a warning.

  • Evidence producers refuse an unqualified KiCad environment (Issue #6962).
    kct readiness and the scripts/boards/ producers now exit 2 and write no
    evidence when kicad-cli is not the pinned image's release (10.0.6) -- e.g.
    Homebrew KiCad 10.0.7 -- or when native DRC reports any
    lib_footprint_issues finding; both used to be recorded silently. Re-run
    with KCT_KICAD_DOCKER=1 to route kicad-cli through CI's pinned image.
    scripts/ci/init_kicad_libraries.py no longer writes into your KiCad
    config: it populates an isolated config home and exports it (to
    $GITHUB_ENV in CI, or via eval "$(python3 scripts/ci/init_kicad_libraries.py --shell)").

  • Code that called a removed router method must switch to the remaining
    API
    (Issue #6984).

    • ViaRules.standard_6layer(): build the rules by hand with
      ViaRules(through_via=ViaDefinition(ViaType.THROUGH, drill_mm=0.3, annular_ring_mm=0.15, start_layer=0, end_layer=5, name="Through")).
    • Corridor helpers: use the remaining RoutingGrid corridor methods,
      set_corridor_preference, clear_all_corridor_preferences and
      get_corridor_cost. There is no per-net clear any more: clear them all
      and re-apply the corridors you want to keep. If you need to know which
      nets have a corridor, record that yourself when you call
      set_corridor_preference.
    • LayerStack.get_layers_adjacent_to_plane and
      get_layers_with_reference_plane: iterate LayerStack.layers and
      filter on layer_type == LayerType.PLANE, plane_net, is_routable
      and reference_plane.
    • Router.set_via_cache_enabled: the via-validity cache is now always
      on, with no off switch. Router.clear_via_cache() still clears it.
  • Rebuild the native router; re-routes may differ (Issue #6994). The C++
    router's BUILD_VERSION is now 60, so run kct build-native after
    upgrading (a v59 extension is rejected and routing falls back to the slower
    Python path). Native A* search radii shrink by one cell where float noise
    used to inflate them (e.g. Default rules on 0.1 / 0.05 / 0.025 mm grids) and
    rip-up considers one more ring of blocking nets, so re-routing an existing
    board can produce different copper.

  • KiCadPCBExporter.generate(), .write() and .generate_trace_segments()
    are removed
    (Issue #7022). They wrote a whole board from a string template
    with footprints that had no pads, so the result could not be routed, checked
    or manufactured. Create the board from the schematic instead (kct create-pcb or schematic-to-PCB sync), then apply the layout to it with
    layout.apply_to_pcb("board.kicad_pcb", "board_placed.kicad_pcb", net_map={...}). update_pcb_placements keeps its signature, now logs a
    warning naming layout refs that are not on the board, and reports the number
    of footprints it actually moved.

  • LEDBlock places its LED at rotation 180 inside the block (Issue #7022),
    so export_placements() reports the LED footprint's rotation 180 degrees
    different from before. Boards laid out with the old orientation shorted the
    LED's anode to its cathode trace.

  • PCB.export_gerbers(..., create_zip=False) on a preset now returns a
    directory
    (Issue #7023). With no create_zip argument the jlcpcb, pcbway
    and oshpark presets still return a ready-to-upload zip; code that passed
    create_zip=False and relied on the zip it got anyway should drop the
    argument. include_drill=False and layers=[...] passed to a preset now
    take effect.

  • kct bom and kct sch summary|validate|preflight|hierarchy exit non-zero on a missing/unparseable schematic (Issue #7028).
    Scripts that rel...

Read more

v0.23.0

Choose a tag to compare

@rjwalters rjwalters released this 15 Oct 11:26
0a0bf62

A large release: 451 entries from about 1,600 commits since 0.22.0.

Read the upgrade notes first. There are 84 of them. kct route and kct check can now fail boards they used to pass:

  • Pads left unconnected on a pour now fail kct route.
  • Signal routing stays off declared plane layers by default.
  • Netclass defaults below the manufacturer's floor are now flagged.
  • The waiver evidence hash format changed, so existing evidence-bound waivers must be re-recorded once (ev1: → ev2:).

Most of the router work targets the lattice engine and the bar for a board to be manufacturable: every net routed, and 0 errors from both kct check and kicad-cli DRC.

  • Diff pairs and match groups on lattice: skew and match-group length are tuned automatically, against the checker's own limits.
  • Pour links: a KiCad-checked completion loop closes pads left unconnected on pours.
  • Inner power plane: it now goes to the power net that feeds the most parts.
  • BGA keep-outs: BGA power balls on the same net get keep-outs.
  • Pad geometry: the clearance code now models roundrect pads exactly.

The demo boards 04, 05 and 06b now route to manufacturable on lattice. 06rf now routes every net and closes every pour link. Its last 3 errors, and those on 07rf, are diff-pair skew that is still being worked on.

New commands and flags:

  • kct ecosystem, kct agent-guide, kct skills install and kct mcp setup --client codex|opencode;
  • kct board-view overlays and kct pcb diff;
  • kct route --checkpoint / --resume;
  • kct route --lint-gate;
  • kct net-status --crossings;
  • evidence-bound waivers and kct check --diff;
  • kct bench external --suite pcbworld.

Several experimental, default-off routing modes also ship: crossing-aware net order, sequential N+1 rip-up, a no-good cache, min-cost-flow escape corridors and bidirectional A*.

The full notes are too long for a GitHub Release. All 84 upgrade notes and every Fixed, Added, Changed and Performance entry are in the 0.23.0 section of CHANGELOG.md.

Install: pip install -U kicad-tools==0.23.0 (PyPI).

v0.22.0

Choose a tag to compare

@rjwalters rjwalters released this 28 Sep 19:52
9e8bc11

Summary

Three new kct check rules — via_under_body and pin1_marker (both on
by default) and the opt-in width_consistency audit — plus native
emission of the different-net SMD-pad clearance floor and a single
clearance resolver for kct route (Epic #5509 Phase 2/3). Read the
upgrade notes:
kct check --strict and native kicad-cli DRC can
newly fail on boards that passed before, and kct route now routes at
the clearance the board's project declares (for example KiCad's stock
0.20 mm Default netclass) instead of letting --mfr or its own target
override it. Also: routing-plan capacity gating (--plan-gate,
net-status --why sidecar), the coupled diff-pair search moved onto the
exact-geometry clearance kernel, pre-release correctness fixes to the
three new rules, and router performance work.

Upgrade notes / behaviour changes

  • kct check --strict can newly fail on boards that passed before.
    Two new categories are on by default: via_under_body (Issue #5736) and
    pin1_marker (Issue #5737). Their findings are warnings, and plain
    kct check exit codes don't change. But --strict, and any CI or ship
    gate that counts warnings, will now fail on:

    • a via under a QFN/DFN/SON/LGA package body (via_under_body)
    • an orientation-sensitive part with no visible silkscreen pin-1 mark
      (pin1_marker_missing / pin1_marker_obscured)

    The repo's own board-04 gate hit this and needed waivers (#5745). To keep
    a gate green without fixing the board, either skip both categories with
    kct check --strict --skip via_under_body,pin1_marker, or add
    per-reference / per-item entries to .kct_waivers.json. Waivers are
    narrower and keep the check active for new parts.

  • kct route now routes at the clearance the board's project
    declares
    (Issues #5645, #5654). The router reads the .kicad_pro and
    .kicad_dru beside the .kicad_pcb — the same set kicad-cli pcb drc
    checks — and treats each declared value as a floor that only ever
    raises the router's 0.15 mm target, never lowers it:

    • an unconditional .kicad_dru clearance rule,
    • the .kicad_pro minimum clearance (min_clearance /
      clearance_min), and
    • the .kicad_pro Default netclass clearance. A project still
      carrying KiCad's stock 0.20 mm Default is now routed at 0.20 mm.
      Projects kct itself creates now write Default at 0.15 mm, so they are
      unaffected.

    In addition, --manufacturer / --mfr no longer overrides a clearance
    the board declares
    ; it only selects which fab floor applies. The
    clearance banner names the layer that won. To keep the previous spacing,
    lower the declared value in the project, or pass --clearance, which
    remains the one deliberate waiver and now warns when it undercuts a
    declared rule (#5656).

  • kicad-cli pcb drc can newly fail boards whose .kicad_dru kct
    regenerates
    (Issue #5059). Generated DRU files (and the shipped JLCPCB
    templates) now carry a native SMD Pad Clearance rule for SMD pads of
    different footprints whenever the profile sets
    min_smd_pad_clearance_mm, so native DRC reports different-net SMD pad
    pairs closer than that floor that it previously passed. KiCad
    10.0.0/10.0.1 never fire the rule, and kct warns when it detects one of
    those versions (#5724).

Fixed

  • Board 04 now silkscreens pin 1 / polarity on U1, U2, D1 and J1
    (Issue #5745). boards/04-stm32-devboard/generate_design.py emitted all
    four of those footprints with no footprint silkscreen graphics at all —
    only the reference text — so pin1_marker_missing fired on every one and
    the four findings had to be waived in output/.kct_waivers.json when
    #5737 landed the rule. Each generator function now emits its stock KiCad
    library artwork: the body outline plus the filled pin-1 triangle for
    SOT-223-3_TabPin2 (Y-mirrored — this board's pad 1 sits at y=+2.3,
    the opposite of the library footprint, so copying the library triangle
    verbatim would have marked pad 3) and for LQFP-48_7x7mm_P0.5mm (four
    corner brackets, transferred as-is — the pad coordinates already match the
    library), the open-sided polarity bracket for LED_0805_2012Metric, and
    the L-shaped pin-1 corner plus body rectangle for
    PinHeader_1x06_P2.54mm_Vertical (shifted from the library's pin-1 origin
    onto this footprint's centered origin). Strokes are 0.15mm, the
    jlcpcb-tier1 silkscreen floor, rather than the library's 0.12mm, and the
    U1/D1/J1 reference designators move clear of the new outlines so
    silk_overlap stays at zero. The four waivers are deleted and the board
    passes the reviewed paid-drill gate (#5009) at strict-0 without them —
    kct check reports 0 errors / 0 warnings / 0 waived, and kicad-cli pcb drc independently reports no silk finding of any kind.

  • The coupled diff-pair search now sees the copper it was routing
    through
    (Issue #5662, Phase 3c of Epic #5509). #4507 summarised the
    defect as "the copper this board fails on is invisible to both by
    construction"
    , and both halves of that are closed here by moving the
    coupled path onto the shared exact-geometry clearance kernel:

    • C++ (CoupledPathfinder::rail_clear, consumer group 7). The gate
      was a lambda inside route(), wrapped in if (grid_.has_fixed_fills()) and consulting fixed fills alone — so on a
      board with no copper pour it ran no rail check at all, and a route
      committed through add_stored_segment / add_stored_via after the
      last blocked-plane re-sync was invisible to it. It is now a named,
      bound method that always runs and measures the candidate rail against
      the grid's stored route geometry with the clearance kernel, through
      the same 2 mm bin index the single-ended refinement path uses. Its two
      rejection reasons also join the #4459 rejection histogram instead of
      pruning the frontier silently.
    • Python (DiffPairRouter's gates, consumer group 8).
      _segment_cells_clear, _span_pad_clear, _segment_pad_clear,
      _span_via_clear, _route_via_clear, _pair_has_physical_overlap,
      _copper_conflicts and find_intra_pair_clearance_violations now
      read the kernel through one shared translation
      (router/clearance_shapes.py) instead of six private predicates.
      _segment_cells_clear gains the missing quadrant — a candidate span
      is screened against the constructor's own unmarked shadow copper —
      and the raster's over-block is re-decided by exact geometry when
      every blocking cell is attributable to copper the grid can name, the
      same authorisation Pathfinder._trace_halo_clear has used since
      #5240. roundrect / oval pads are now measured as the copper they
      are rather than as their enclosing rectangle.
    • The oracle rows harden. Groups 7 and 8 stop being report-only in
      the over-rejection direction: refusing a candidate kicad-cli calls
      clean is a test failure
      (test_migrated_consumer_never_over_rejects). Group 7 was the epic's
      one unexposed consumer; binding rail_clear retires
      NOT_MEASURED_REASONS entirely, so all nineteen groups are measured.
    • No rule value, .kicad_dru, or A* heuristic term is touched (Epic
      #5509 scope guards #1 and #2); rule resolution stays with each
      consumer, the kernel answers only gaps.
  • Coupled differential-pair search now refines dynamic route-halo rejection
    with physical clearance
    (part of Issue #5410). A dynamic route halo is a
    conservative acceleration structure: committed copper is dilated to whole
    grid cells, so a foreign route's halo covers candidates whose actual copper
    and drill gaps satisfy the effective rules. PR #5425 taught the per-net A*
    (both backends) to measure that geometry before rejecting such a candidate;
    the coupled joint-state search — the engine that routes every
    differential pair — still answered on the raster alone, so every pair after
    the first refused legal steps beside already-committed copper. Both coupled
    backends now apply the identical check, over the swept trace step (not just
    its endpoint) and the through-via envelope. Hard or unverifiable blockage —
    pad metal, static halos, keepouts, reservations, hidden overlapping owners,
    missing geometry — still rejects, and the refinement is fail-closed:
    dormant without the net-name map it needs to resolve pairwise
    (HV-isolation) requirements, and, on the C++ side, dormant unless that
    grid's own cross-domain widening was installed and the refiner is armed.
    The refinement re-measures with the candidate's net-class trace width,
    clearance and via size on both backends — re-measuring with the global
    DesignRules scalar would have waived clearance the raster halo (dilated
    with the net-class value) had enforced, which is the #5673 under-blocking
    direction. Native ABI 42 → 44.

  • Qualify KiCad 10.0.6 for the native mask-to-copper gate (Issue #5678).
    The CI runner's floating kicad/kicad:10.0 tag moved from 10.0.5 to 10.0.6,
    and both hard-coded 10.0.5 version pins — scripts/ci/check_mask_copper_native.py's
    gate and validate/mask_copper_geometry.py's native-attribution check —
    rejected it, failing the Test job's mask step (which also skipped every
    later step in that job, including Run tests). The two pins are now one
    shared QUALIFIED_NATIVE_VERSIONS = ("10.0.5", "10.0.6") in
    validate/mask_copper_geometry.py, imported by the CI gate so the pair can
    no longer drift apart; an unqualified build still fails loudly, now naming
    the versions it did see.
    10.0.6 was measured, not assumed, before being admitted: the full native
    suite was run in both kicad/kicad@sha256:182c8005… (10.0.5) and
    kicad/kicad@sha256:18693567… (10.0.6), each reporting 32 passed, no skips, and the resulting mask/copper geometry was compared across the two —
    every exported layer polygon and every ...

Read more