VOGONS


dgVodooo 2.8.x and related WIP versions

Topic actions

Reply 200 of 204, by johnnb

User metadata
Rank Newbie
Rank
Newbie

Dear guys, how are you I have one question if possible of course, about dgvoodoo 2

Reply 201 of 204, by Deffnator

User metadata
Rank Member
Rank
Member
johnnb wrote on 2026-01-23, 18:23:

Dear guys, how are you I have one question if possible of course, about dgvoodoo 2

Just tag dege or someone else with knowledge will answer

Reply 202 of 204, by Squall Leonhart

User metadata
Rank Member
Rank
Member

DgVoodoo leaks d3d12 resources when presentation and rendering are not on the same GPU (muxless switchable graphics behavior) on
intel + Geforce configurations.

Reply 203 of 204, by Dege

User metadata
Rank l33t
Rank
l33t
Squall Leonhart wrote on 2026-02-01, 09:13:

DgVoodoo leaks d3d12 resources when presentation and rendering are not on the same GPU (muxless switchable graphics behavior) on
intel + Geforce configurations.

No, it does not. If you mean that guru3dforums (or whatever) thread, that was actually me. But I have a plain desktop machine so it's not even a muxless case, it's something that they just imagined for themselves. But even with a muxless setup, dgVoodoo runs on the same path so it should not matter.

Reply 204 of 204, by TianQI233

User metadata
Rank Newbie
Rank
Newbie

Bug: Window shifted down in fake fullscreen on non-integer DPI scaling (16:10 display)
Hi Dege,
I found a window positioning bug with dgVoodoo 2.86.4 D3D9 wrapper.
Environment:

Windows 11 24H2
Display: 2560×1600 (16:10)
DPI scaling: 150%
Game: Touhou 16 (Hidden Star in Four Seasons), native D3D9, using thcrap launcher

Symptom:
In fake fullscreen mode, the entire game image is shifted downward by ~30-40px. The bottom of the game is clipped below the screen edge, and the top portion appears vertically stretched to fill the gap — as if the client area is being rendered at an offset and the display output is scaled to compensate.
Reproduction:

2560×1600 @ 150% → bug occurs
2560×1600 @ 200% → no issue (DPI virtualization masks it)

Hint on location:
The issue seems to be in the windowed-mode window creation path. When the window is created with WS_CAPTION style, AdjustWindowRect expands the rect to include the title bar. The centering logic then computes a negative top coordinate to hide it above the screen. But there is a clipping check that detects the window extends above the monitor top and calls OffsetRect to push it back down — this defeats the purpose of the negative offset and shifts the client area down by the title bar height.
On 200% scaling, Windows DPI virtualization reports a smaller virtual resolution so the offset is small enough to go unnoticed. On 150% (non-integer), the wrapper operates closer to physical coordinates and the shift becomes clearly visible.
Thanks for the amazing work on dgVoodoo!