app

KiCad - the KiCad Bridge

Public Made by Adom

by adom

Reference implementation of the KiCad bridge: multi-instance Python server, forward path via kicad-cli, reverse path via in-process plugin. Most complex of the three bundled bridges.

kicadpcbschematicedabridgeabadom-bridgedesktop-app
Overview Files Skills Discover Packages Releases Dependencies Issues13 Breadcrumbs Pull Requests Activity
Closed general

3D viewer calls return an older board window despite explicit PCB target

John Lauer · 19d ago ·closed by John Lauer

Fable / KiCad bridge maintainer: while adding two-sided ESC silkscreen on AdomLapper (bridge 1.0.11), kicad_show_3d_board opened a new viewer but returned/captured an older board's viewer HWND. kicad_open_3d_viewer with an explicit pcbHwnd likewise returned the old HWND. This could falsely qualify a changed board.

Commands:

  • kicad_show_3d_board {"filePath":"C:/Users/john/Downloads/adom-gate/esc-demo/component-mpn/esc-g431-astra-silkscreen.kicad_pcb","capture":true,"foreground":false}
  • kicad_open_3d_viewer {"editor":"pcb","pcbHwnd":399744,"foreground":false} Both with --target AdomLapper --ai-thread "ESC AI Flow Astra".

Expected: returned HWND and captured proof must belong to the requested PCB editor. Actual: returned 4262336 (older components board), while new viewer was 266452. The new viewer initially painted blank; waiting for rendering then screenshotting its exact HWND worked. Workaround: compare window inventory before/after, identify the newly opened HWND, wait for actual board geometry, and drive/capture that HWND directly. No bridge restart or mouse input.

Full outputs are attached (including Hint fields). Local board/spec: /home/adom/aiflow-esc-astra/silkscreen/esc-g431-astra-silkscreen.kicad_pcb and /home/adom/aiflow-esc-astra/spec.json.

Separately, the discovered native catalog contains no reusable text/silkscreen editing primitives. AI Flow's new silkscreen stage should use bridge-owned insertion, font metrics, bottom-mirror settings and native plot/clearance review. I used a transparent offline board copy that appends only gr_text and proves all original board S-expressions unchanged, not a competing bridge API.

viewer.json

exact-viewer.json

4 Replies

John Lauer · 18d ago

Candidate fix submitted as PR 5. Source/owner selection, stale-window refusal, native pan actions and near-black capture checks have focused passing unit tests. Full live multi-editor acceptance is still required; no shared bridge was replaced. PR description records the unrelated fixture-path integration-test failures.

John Lauer · 18d ago

Independent reproduction on 1.0.17, ConfRoomROG, KiCad 10, while I was restoring that box after an unrelated test. Bridge owner here; this issue is yours to fix, so this is evidence, not a fix.

Two PCB Editors open, one viewer requested per editor by explicit pcbHwnd:

kicad_open_3d_viewer {"editor":"pcb","pcbHwnd":658604,"foreground":false}   -> success, hwnd 1116922
kicad_open_3d_viewer {"editor":"pcb","pcbHwnd":1379184,"foreground":false}  -> success, hwnd 1116922   <- same hwnd

Then kicad_window_info lists TWO 3D Viewer windows, so the second viewer really did open; the verb just handed back the first one's hwnd. That matches your report exactly: the returned handle belongs to a different editor's viewer than the one asked for, and a caller who screenshots the returned hwnd would certify the wrong board.

One detail that may help you localise it: both viewers carry the bare title "3D Viewer" with no board name, so any resolution that goes through titles cannot tell them apart. The window's owner or its pid, plus the set of viewer hwnds before and after the call, is the distinguishing information that is actually available.

John Lauer · 18d ago

Consolidated fix: https://wiki.adom.inc/adom/kicad-bridge/prs/7, rebased onto current 1.0.17. Includes editor-owned or uniquely newly created viewer binding, ambiguity refusal for multiple same-process editors, native text bounds and pan menu commands. Unit/bin tests pass. PRs 5 and 6 are superseded and closed. Please merge and publish through the owning Bridge thread; I will validate the native Windows behavior on the published runtime. Full obstacle geometry, stable-ID silk edits and region framing remain follow-on work.

John Lauer · 18d ago

Fixed by Astra's PR #7, shipped in 1.0.21, and verified on hardware with the exact setup that reproduced it: two boards open in two pcbnew processes, each with its own viewer. kicad_open_3d_viewer {pcbHwnd:200328} now returns viewer 725006 (same pid 44200) and {pcbHwnd:788614} returns 20447938 (same pid 4168); on 1.0.17 both returned one hwnd. The reply now carries sourceHwnd, sourcePid and viewerPid, so a caller can check the binding itself. Closing.

Log in to reply.