D2D coordinates not matching screen coordinates - c++

Question

I am using D2D with D3D11. I have some code that uses GetCursorpos() from the windows API which is then converted to client coordinates and then draws a small circle at this position using D2D FillEllipse(). The screen to client coordinates work perfectly but for some reason D2D draws the circle a small distance from the expected location (tens of pixels) as if the coordinate had been scaled by a small factor so that the error gets larger as the circle is drawn further from (0, 0).
I noticed changing the dpi for the D2D1_RENDER_TARGET_PROPERTIES affects this 'scaling' so I suspect the problem has something to do with dpi. This is the code for creating the D2D render target from the DXGI surface I obtained from the swapchain in my D3D11 code.
// Create render target
float dpiX, dpiY;
this->factory->GetDesktopDpi(&dpiX, &dpiY);
D2D1_RENDER_TARGET_PROPERTIES rtDesc = D2D1::RenderTargetProperties(
D2D1_RENDER_TARGET_TYPE_HARDWARE,
D2D1::PixelFormat(DXGI_FORMAT_UNKNOWN, D2D1_ALPHA_MODE_PREMULTIPLIED),
dpiX,
dpiY
);
AssertHResult(this->factory->CreateDxgiSurfaceRenderTarget(
surface.Get(),
&rtDesc,
&this->renderTarget
), "Failed to create D2D render target");
Here, dpiX and dpiY become 96 which I notice is also the constant that GetDpiForWindow() from the windows API returns when it is not dpi aware.
I want to know how I can fix my code so that it will draw the circle at the position given by GetCursorPos().
More relevant code:
Driver code
Vector3f cursPos = input.GetCursorPos();
DrawCircle(Colour::Green, cursPos.x, cursPos.y, 3/*radius*/);
Input
POINT pt{};
::GetCursorPos(&pt);
// Convert from screen pixels to client pixels
return ConvertPixelSpace(this->hWnd, (float)pt.x, (float)pt.x, PixelSpace::Screen, PixelSpace::Client);
Direct2D
void DrawCircle(const Colour& c, float centreX, float centreY, float radius, PixelSpace ps)
{
Vector3f centre = ConvertPixelSpace(this->gfx.hWnd, centreX, centreY, ps, PixelSpace::Client);
centreX = centre.x;
centreY = centre.y;
D2D1_ELLIPSE el{};
el.point.x = centreX;
el.point.y = centreY;
el.radiusX = radius;
el.radiusY = radius;
auto brush = this->CreateBrush(c);
this->renderTarget->FillEllipse(
&el,
brush.Get()
);
}
PixelSpace Conversion
Vector3f ConvertPixelSpace(HWND hWnd, float x, float y, PixelSpace curSpace, PixelSpace newSpace)
{
RECT rc = GetClientRectOfWindow(hWnd);
struct
{
float top, left, width, height;
} rectf;
rectf.top = static_cast<float>(rc.top);
rectf.left = static_cast<float>(rc.left);
rectf.width = static_cast<float>(rc.right - rc.left);
rectf.height = static_cast<float>(rc.bottom - rc.top);
// Convert to client space
if (curSpace == PixelSpace::Screen)
{
x -= rectf.left;
y -= rectf.top;
}
// Convert to new space
if (newSpace == PixelSpace::Screen)
{
x += rectf.left;
y += rectf.top;
}
return Vector3f(x, y);
}
RECT GetClientRectOfWindow(HWND hWnd)
{
RECT rc;
::GetClientRect(hWnd, &rc);
// Pretty sure these are valid casts.
// rc.top is stored directly after rc.left and this forms a POINT struct
ClientToScreen(hWnd, reinterpret_cast<POINT*>(&rc.left));
ClientToScreen(hWnd, reinterpret_cast<POINT*>(&rc.right));
return rc;
}

Answer 1

The problem was I was creating the D3D11 swapchain with the window area instead of the client area.
RECT rect{};
GetWindowRect(hWnd, &rect); // !!! This should be GetClientRect()
this->width = rect.right - rect.left;
this->height = rect.bottom - rect.top;
DXGI_SWAP_CHAIN_DESC scDesc{};
scDesc.BufferDesc.Width = width;
scDesc.BufferDesc.Height = height;
//...