# #4841 fixture matrix v3: observed outcomes (supersedes the lead's predicted expectations in matrix v2)

Before = 689763a603, After = 6fd7ae506e. Linux Electron 44.9.3, Xorg dummy two outputs, openbox. Captures: https://gh-file-drop-api-prod-mi5fy3sowv63ufte.pinglabs.workers.dev/f/b9c5c72594d12626/evidence.md

| Row | Observed before | Observed after | Status |
|---|---|---|---|
| L01 control | normal on display 2 | same | unchanged, as intended |
| L02 move then maximize (reporter's order) | maximized on display 2 (Linux WM path reports maximized normal bounds) | same | unchanged on Linux; the macOS path is cannot-verify |
| L03 app-restored maximize moved to display 2 | maximized on display 1 | maximized on display 2 | fixed |
| L03s straddling normal bounds | the WM replaces straddling normal bounds on maximize, so the condition is not reachable on Linux | same | unit test only |
| L04 smaller display 2 | maximized on display 2 (prediction "display 1" was wrong: on this path Linux normal bounds already sit on display 2) | same | unchanged on Linux; the clamp is unit-tested at 1280x720 |
| L09 fullscreen after app-restored maximize | display 1 | display 2 (fullscreen is not restored on either side) | fixed |
| L12 minimized quit, settled save | normal on display 2 (maximized flag lost) | maximized on display 2 | fixed: maximized state survives a minimized quit |
| L12b minimized quit within 250 ms | normal on display 2 (maximized flag lost) | maximized on display 2 | fixed |
| L13 normal move, minimized within 250 ms | normal at 2300,150 on display 2 | same | no regression (#3851 path) |
| L10 unmaximize after the transfer | normal on display 1 (native restore frame) | same | unchanged: the patch changes persisted bounds, not Electron's native restore frame |
| P01 macOS | — | — | cannot-verify macos |
| P02 Windows | — | — | cannot-verify windows |