erik/acdream
1
0
Fork 0
Code Issues Pull requests Projects Releases 6 Packages Wiki Activity Actions

fix #374: open dropdown popups get first claim on pointer routing

Browse source 
Campaign OP gate 2 root cause: UiElement.HitTest walks siblings
front-to-back by z-order, so an OPEN UiMenu's extended button+popup
hit-test union was never consulted when a LATER sibling's rect overlapped
the popup area — on the Config tab every dropdown has rows below it, so
Resolution-item clicks toggled the Full Screen / VSync rows underneath
(the gate session's persisted fullscreen/vsync flips were exactly those
stolen clicks). Latent since UiMenu existed; vendor/chat menus only
worked by z-order luck.

Fix: UiMenu's open/close now registers with UiRoot (SetActivePopup /
ClearActivePopup); a registered popup gets FIRST claim on mouse-down,
scroll, and hover routing; a press outside a live popup dismisses it and
is SWALLOWED (the dismissing click must not act on what sat underneath);
hidden/detached owners self-heal the registration on the next pointer
event. UiMenu gains the IsOpen seam and a single SetOpen writer.

Also in this commit, from the same investigation:
- SilkRuntimeDisplayWindowTarget.Apply documents the fullscreen half
  honestly: IViewProperties.VideoMode is READ-ONLY, so a resolution pick
  while fullscreen cannot switch the display mode through Silk's
  abstract API — split out as #376 (native glfwSetWindowMonitor port)
  rather than half-shipping untested native interop at a gate tail.
- Gate script §OP6 step 8 re-scoped: test resolution in WINDOWED mode.

Regressed by tests/AcDream.App.Tests/UI/UiMenuPopupRoutingTests.cs —
4 tests driving the real UiRoot input path on a mounted overlapping
tree, with an in-test overlap CONTROL click so the popup assertions
cannot pass vacuously (the #372 lesson: only mount+drive-input tests
catch this class; every fixture-conformance test stayed green through
this bug).

Full Release suite: 13,081 passed / 4 skipped / 0 failed.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Erik 2026-08-11 15:17:17 +02:00
parent 07c0c2c7b9
commit 355c86a6f6
6 changed files with 388 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/AcDream.App/Settings/RuntimeSettingsTargets.cs
View file

@ -56,8 +56,9 @@ internal sealed class SilkRuntimeDisplayWindowTarget : IRuntimeDisplayWindowTarg
public void Apply(DisplaySettings display)
{
ArgumentNullException.ThrowIfNull(display);
if (TryParseResolution(display.Resolution, out int width, out int height)
&& (_window.Size.X != width || _window.Size.Y != height))
bool haveResolution =
TryParseResolution(display.Resolution, out int width, out int height);
if (haveResolution && (_window.Size.X != width || _window.Size.Y != height))
{
_window.Size = new Vector2D<int>(width, height);
}
@ -65,6 +66,16 @@ internal sealed class SilkRuntimeDisplayWindowTarget : IRuntimeDisplayWindowTarg
WindowState desired = display.Fullscreen
? WindowState.Fullscreen
: WindowState.Normal;
// #374 investigation note: while FULLSCREEN, the visible resolution
// is the display's video mode, and Silk's abstract windowing API
// cannot change it (IViewProperties.VideoMode is read-only; Silk
// fullscreen is desktop-mode borderless). The Size write above is
// therefore only visible in windowed mode — a resolution pick while
// fullscreen changes what a later return to windowed restores, not
// the fullscreen mode itself. Retail's own fullscreen resolution
// switch (Device::ForceDisplayResolution) needs a native
// glfwSetWindowMonitor port — issue #376.
if (_window.WindowState != desired)
_window.WindowState = desired;
}