[yacineolivier-maker]

A window exactly covering the display is promoted by Windows to a direct /
independent flip. OpenGL swaps issued from the render thread then stop
reaching the screen on NVIDIA drivers: the image freezes while the game keeps
running, and it comes back as soon as anything forces composition (taskbar,
another window, a tooltip). The borderless window is now one row taller than
the display (the extra row is off-screen), so it always stays composed.
test_window_fullscreen checks the new size.

Seen with R4 on an RTX 3060 (driver 576.57), Windows 11, render thread on.

Verification

• Game-agnostic: no title check or game address in the framework; behaviour is unchanged unless borderless fullscreen is used on Windows.
• Build: R4 (tetrisgm/RidgeRacerType4Recomp master 737827c, USA config, its committed generated C) built against this branch alone: cmake-clang-v1 1.0.14 (clang, Ninja, Release), Windows 11.
• Runtime: exercised in R4's PAL build (SCES-01706) with all of our framework branches combined on current master fe8d937d: boot, menus, a full Time Attack lap. The same changes on the previous base (379565a6) went through several days of play-testing: intro movie, menus, races, VS split screen, attract demo. Windows 11, i5-12400F, RTX 3060, OpenGL renderer; Vulkan and the software renderer were not tested.
• Test: test_window_fullscreen updated for the new size; not run here (it is the opt-in desktop test, PSX_WINDOW_FULLSCREEN_TEST). Before/after checked by hand in R4: frozen image with the render thread on before, none after.
• Pin / regen: runtime only, no regen. A game consumes it with a pin bump.

Game side: tetrisgm/RidgeRacerType4Recomp#49 (PAL / Europe support) and tetrisgm/RidgeRacerType4Recomp#50 (race HUD, car effects, lighting; overview in tetrisgm/RidgeRacerType4Recomp#51)

I cannot follow up on review comments, sorry. Maintainers are welcome to edit, split, take over or close this.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Rkyma8rbyPiccs9n2bkyG6

[mstan]

Reviewed head 20a9312.

The reported NVIDIA render-thread freeze merits a default correctness fix. This patch applies a one-pixel oversized window to every Windows borderless session, including software/Vulkan and configurations that do not use the affected GL render thread. That changes drawable height/aspect and can cross a DPI/monitor boundary; updating a test to require h+1 only asserts the workaround. Please scope the affected presentation path using the effective renderer/window/thread state, or provide evidence that the broader borderless policy is desired. Validate actual captions/window bounds, multi-monitor and non-100% DPI behavior, fullscreen exit/re-entry, and capture the real window to establish continuing presents. This is not a reason to put the correctness repair behind an enhancement opt-in. Coordinate with #723 rather than layering two unqualified platform-wide pacing workarounds. Tracking: beads-8zz2.

[mstan]

Rebased cleanly onto current master and pushed 1697e91. The dummy-SDL geometry/restore test passes, but that is not evidence that the one-extra-row workaround prevents independent flip on an affected Windows desktop.

Before treating this as ready, please provide a reproducible affected driver/DPI/monitor case and DWM/PresentMon evidence for the proposed scope/default. The available host has a different NVIDIA driver from the reported reproducer. A universal fullscreen-size policy is a design/qualification decision; I have not silently changed it or claimed a hidden-window test proves compositor behavior.