app

KiCad - the KiCad Bridge

Public Made by Adom

by adom

Drive your own KiCad from an AI, in the background: open every editor, read the whole board as data (the EDA Bridge Standard board model), stack-up, part heights, netlists, place, route, pour, DRC and exports.

kicadpcbschematicedabridgeabadom-bridgedesktop-app
Overview Files Skills Discover Packages Releases Dependencies Issues Breadcrumbs Pull Requests1 Activity
Closed general

Reminder for John: connect arav-rog to the desk monitor so we have a multi-monitor test box

John Lauer · ·closed by John Lauer
reminder

@John Lauer reminder, at your request: connect arav-rog to the monitor on your desk, so Adom has a test box with two displays.

Why it matters. Every window-placement fix we ship is currently verified on single-screen boxes. #48 is the live example: the placement memory is proven on hardware (parked a window, killed KiCad so it could not save its own geometry, and the bridge still put the window back at 300,200 1200x800), but the actual complaint is "I drag KiCad to my external monitor and it comes back on the laptop screen", and nobody at Adom can reproduce that today. The same gap applies to anything else that reasons about monitors: the idle healer's on-screen check, the tour and demo window layout, and full-screen capture.

What the box would let us test, in minutes, once it has two displays:

  1. Drag a PCB Editor to the second monitor, then kicad_rescan_libraries {force:true, confirmRestart:true}, and confirm it comes back on that monitor with placementRestored[].moved true.
  2. Undock or unplug the second monitor, reopen the editor, and confirm the bridge REFUSES to place a window on a display that is no longer there.
  3. Leave a window on the second monitor for a few minutes and confirm the 20 s idle healer leaves it alone (the 0.9.318 fix, ported to the native build but never seen on real hardware).

One standing rule that does not change: arav-rog is Caleb's machine and no AI thread touches it. This is a hardware change for you to make, and the box only becomes a test target if you say it is one.

Filed as a reminder rather than kept in a chat thread so it does not evaporate when the session ends.

2 Replies

John Lauer ·

Planned for tomorrow, Saturday 2026-09-19 (John). Once the second display is connected, say so here and I will run the three tests above and post the results.

John Lauer ·

John connected the second monitor and designated arav-rog for this, so all three tests ran today on bridge 1.0.33, KiCad 10.0.3. The box has a 2560x1600 primary at 150% and a 2560x1440 second display at 100%, placed to the LEFT (x from -2560), so this also covered negative coordinates and mixed DPI.

Test Result
1. Restart restore on the second monitor Pass. Easy case: editor parked at -2300,300 1600x1000, kicad_rescan_libraries {force, confirmRestart} closed gracefully, the reopened editor was at -2300,300 (moved:false, KiCad's own saved geometry agreed). Hard case: closed on the primary so KiCad's config said primary, reopened, moved to -2200,400 1700x950 on monitor 2, force-killed, reopened. Reply: placementRestored {moved:true, to:{x:-2200,y:400,width:1700,height:950}}, and the real window rect read back as left -2200, top 400, 1700x950.
2. Refuse a display that is gone Pass, simulated. Nobody can unplug a monitor remotely, so the memory file was pointed at x 6000 (no display there). Reply: moved:false, why:"the remembered rect is on no monitor that is connected now", and the window stayed where KiCad put it (200,150 on the primary). A physical unplug is still worth one look when someone is at the desk.
3. Idle window left alone Pass. The editor sat on monitor 2 for two minutes with the bridge idle and its rect did not change. Worth knowing: the native build has no idle window healer at all (that was the Python build's), so nothing in it moves a window nobody asked about.

One bug found and fixed, 1.0.34 (insiders). arav-rog's placement file was still in the flat format the Python build wrote ("pcb editor": {x, y, w, h}). The native loader read that as empty, so its first save replaced the file and dropped every other window kind. 1.0.34 reads the old format too, with a unit test. Verified live on arav-rog with the original file: the editor went to the remembered 0,0 2560x1368 (moved:true), and after a close the rewritten file holds all six window kinds.

arav-rog is left as found: KiCad closed, no dialogs, the original placements in the file. The test fixture is in C:/Users/arav/Downloads/kicad-bridge-107.

Closing: the reminder is done and the tests it was for have run.

John Lauer closed this ·

Log in to reply.