fix(UI): correct ImGui scaling for borderless mode

[zndxcvbn]

Fixes #1319.

Summary by CodeRabbit

• New Features

  • UI now updates input from the window before each UI frame to ensure consistent input handling.

• Bug Fixes

  • Improved high-DPI/screen-scaling support for UI rendering.
  • Enhanced mouse input handling with proper scaling and coordinate conversion for accurate, responsive cursor positioning across displays.

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

Walkthrough

Queries DXGI swap chain for the target window, computes/display-scale metrics, updates ImGui io.DisplaySize and feeds scaled mouse coordinates via Util::UpdateImGuiInput(HWND), and calls that utility from OverlayRenderer immediately before ImGui::NewFrame().

Changes

Cohort / File(s)	Summary
Screen scaling utilities src/Utils/UI.h, src/Utils/UI.cpp	Adds DXGI include and internal state for screen scale and display size. Implements Util::RefreshScreenScale(HWND) and Util::UpdateImGuiInput(HWND) to compute display size/scale from swap chain and push scaled DisplaySize and mouse coordinates to ImGui.
ImGui frame integration src/Menu/OverlayRenderer.cpp	Queries swap chain for OutputWindow HWND (if available) and calls Util::UpdateImGuiInput(hwnd) immediately before ImGui::NewFrame(). Adds #include <dxgi.h> usage context.

Sequence Diagram

sequenceDiagram
    participant Overlay as OverlayRenderer::InitializeImGuiFrame
    participant Util as Util::UpdateImGuiInput
    participant Init as Util::RefreshScreenScale
    participant DXGI as DXGI SwapChain
    participant Win as WinAPI (GetCursorPos / ScreenToClient)
    participant ImGui as ImGui

    Overlay->>Util: call UpdateImGuiInput(hwnd)
    Util->>Init: ensure refresh (RefreshScreenScale(hwnd))
    Init->>DXGI: query swap chain desc / OutputWindow
    DXGI-->>Init: return display dimensions / HWND
    Init->>Init: compute g_screenScaleRatio & g_displaySize
    Init-->>Util: return
    Util->>ImGui: set io.DisplaySize (g_displaySize)
    Win->>Util: GetCursorPos -> ScreenToClient (client pos)
    Util->>ImGui: AddMousePosEvent(scaled client pos)
    Util-->>Overlay: return
    Overlay->>ImGui: ImGui::NewFrame()

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Suggested reviewers

• doodlum
• alandtse
• SkrubbySkrubInAShrub

Poem

🐇 I hopped through DXGI's light,
Measured pixels left and right.
I scaled the cursor, set the view—
Frames align and menus too.
🥕✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 16.67% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: fixing ImGui scaling in borderless mode, which directly addresses the linked issue about ImGui window size mismatches in borderless mode with custom resolutions.
Linked Issues check	✅ Passed	The code changes directly address issue #1319 by implementing display scaling awareness via RefreshScreenScale() and UpdateImGuiInput() to handle borderless mode with custom resolution mismatches.
Out of Scope Changes check	✅ Passed	All changes are focused on fixing the ImGui scaling issue in borderless mode: adding DXGI awareness, screen scaling calculations, and input coordinate transformation aligned with the issue requirements.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

Actionable Suggestions

• Hair Specular (doodlum): Needs version bump to 1-1-0
• Terrain Shadows (doodlum): Needs version bump to 1-1-0
• Volumetric Shadows (doodlum): New ini added (v2-0-0)

[coderabbitai]

Actionable comments posted: 3

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@src/Menu/OverlayRenderer.cpp`:
- Around line 126-138: GetClientRect(desc.OutputWindow, &rect) can fail and
currently there is no else branch so screenScaleRatio may remain stale; update
the code around the GetClientRect call in OverlayRenderer.cpp (the block that
computes screenScaleRatio) to set screenScaleRatio = ImVec2(1.0f, 1.0f) when
GetClientRect returns FALSE (or initialize screenScaleRatio to that fallback
before the call), ensuring the fallback is applied both when GetClientRect fails
and when width/height are zero in the existing inner else.
- Around line 119-124: The call to globals::d3d::swapChain->GetDesc(&desc) is
not checking the returned HRESULT, so a failure leaves desc zeroed and sets
io.DisplaySize to (0,0); modify the code around the GetDesc usage in the ImGui
display sizing block (and the duplicate call in UpdateMousePosition) to capture
the HRESULT, only use desc.BufferDesc.Width/Height if SUCCEEDED(hr), and
otherwise fall back to the existing GetClientRect path or preserve the previous
io.DisplaySize; ensure both places (the DisplaySize assignment and the
UpdateMousePosition GetDesc call) perform the same HRESULT check and fallback
behavior.
- Around line 141-154: UpdateMousePosition currently calls
globals::d3d::swapChain->GetDesc() again just to read desc.OutputWindow and
reads screenScaleRatio which can be uninitialized; instead cache the HWND
obtained in UpdateDisplayMetrics (called from InitializeImGuiFrame) and pass it
into UpdateMousePosition so you can remove the extra GetDesc() COM call, and
ensure screenScaleRatio (or a local scale variable) is initialized to a safe
default (e.g., {1.0f, 1.0f}) at declaration or set unconditionally after
GetClientRect() so UpdateMousePosition and UpdateDisplayMetrics always see a
valid value; update usages of UpdateMousePosition, UpdateDisplayMetrics, and any
callers in InitializeImGuiFrame to accept the cached HWND and rely on the
initialized screenScaleRatio.

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

src/Menu/OverlayRenderer.h (1)

48-59: ⚠️ Potential issue | 🔴 Critical

UpdateDisplayMetrics() and UpdateMousePosition() must be declared in the class definition

Both methods are defined in OverlayRenderer.cpp as OverlayRenderer::UpdateDisplayMetrics() and OverlayRenderer::UpdateMousePosition(), but neither is declared in the private: section here. C++ requires an out-of-line member function definition to have a matching in-class declaration; without it this is a hard compiler error.

🐛 Proposed fix

 	static void FinalizeImGuiFrame();
+	static void UpdateDisplayMetrics();
+	static void UpdateMousePosition();
 	static ImVec2 screenScaleRatio;

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/Menu/OverlayRenderer.h` around lines 48 - 59, The class OverlayRenderer
is missing in-class declarations for the out-of-line methods defined in
OverlayRenderer.cpp; add matching private static declarations for
UpdateDisplayMetrics() and UpdateMousePosition() (e.g., static void
UpdateDisplayMetrics(); static void UpdateMousePosition();) alongside the other
private static methods so the out-of-class definitions
(OverlayRenderer::UpdateDisplayMetrics and OverlayRenderer::UpdateMousePosition)
have corresponding declarations.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Outside diff comments:
In `@src/Menu/OverlayRenderer.h`:
- Around line 48-59: The class OverlayRenderer is missing in-class declarations
for the out-of-line methods defined in OverlayRenderer.cpp; add matching private
static declarations for UpdateDisplayMetrics() and UpdateMousePosition() (e.g.,
static void UpdateDisplayMetrics(); static void UpdateMousePosition();)
alongside the other private static methods so the out-of-class definitions
(OverlayRenderer::UpdateDisplayMetrics and OverlayRenderer::UpdateMousePosition)
have corresponding declarations.

---

Duplicate comments:
In `@src/Menu/OverlayRenderer.cpp`:
- Around line 128-140: GetClientRect can fail after the first frame which leaves
screenScaleRatio with a stale value; update the block in OverlayRenderer.cpp
where GetClientRect(desc.OutputWindow, &rect) is called so that if GetClientRect
returns FALSE you explicitly reset screenScaleRatio to ImVec2(1.0f, 1.0f) (or
the safe fallback), ensuring the variable is not left from a previous frame;
locate the code around the existing GetClientRect call and add an else branch
that assigns the fallback to screenScaleRatio, keeping the existing inner check
for windowWidth/windowHeight intact.
- Around line 121-126: Check the HRESULT returned by
globals::d3d::swapChain->GetDesc(&desc) before using desc; if GetDesc fails, do
not set io.DisplaySize from the zero-initialized desc. Update the code around
the GetDesc call (reference globals::d3d::swapChain, GetDesc, desc,
io.DisplaySize, ImVec2) to capture the HRESULT, handle failure by
logging/returning early or preserving a previously stored display size/fallback
values, and only assign io.DisplaySize when GetDesc succeeds.
- Around line 147-148: UpdateMousePosition currently calls
globals::d3d::swapChain->GetDesc() just to read desc.OutputWindow and ignores
the HRESULT; instead, have InitializeImGuiFrame obtain and validate the HWND via
UpdateDisplayMetrics and then pass that HWND into UpdateMousePosition (or store
it in an additional static member) so UpdateMousePosition no longer calls
swapChain->GetDesc()/OutputWindow or repeats the COM call and error path; update
signatures/usages for InitializeImGuiFrame, UpdateMousePosition, and
UpdateDisplayMetrics accordingly and remove the redundant GetDesc() call and
unchecked HRESULT in UpdateMousePosition.

[coderabbitai]

Actionable comments posted: 3

🧹 Nitpick comments (1)

src/Utils/UI.h (1)

52-52: Consider adding a doc comment for UpdateImGuiInput.

Every other public function in namespace Util has a Doxygen-style doc block. This declaration has none, which breaks the established documentation pattern.

📝 Suggested doc comment

+	/**
+	 * `@brief` Updates ImGui IO DisplaySize and mouse position with borderless-mode scaling applied.
+	 * Initializes the screen-scale ratio once (swap-chain dimensions vs window client area)
+	 * and feeds scaled cursor coordinates to ImGui each frame.
+	 * `@param` hwnd The game window handle.
+	 */
 	void UpdateImGuiInput(HWND hwnd);

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/Utils/UI.h` at line 52, Add a Doxygen-style doc comment above the
declaration of UpdateImGuiInput in namespace Util describing its purpose,
parameters and behavior (e.g., what hwnd is and how the function updates ImGui
input state), matching the style used by the other public functions in the
header so documentation is consistent; place the comment immediately above the
line "void UpdateImGuiInput(HWND hwnd);" and include `@param` for hwnd and a brief
`@brief` sentence about side effects.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@src/Utils/UI.cpp`:
- Around line 55-57: The code sets g_scaleInitialized once and returns early,
which leaves g_screenScaleRatio stale after a swap chain resize; update the
logic so the scale is recomputed when the swap chain is recreated: either
clear/reset g_scaleInitialized during the swap chain recreation path (where the
swap chain or its size is updated) or modify the getter (the function containing
the early return using g_scaleInitialized) to do a cheap staleness check
(compare current backbuffer/viewport dimensions to cached values) and recompute
g_screenScaleRatio if dimensions changed; reference g_scaleInitialized,
g_screenScaleRatio and the swap chain recreate code path when implementing the
fix.
- Around line 53-71: Move the g_scaleInitialized = true assignment inside the
success branch so initialization only marks complete when the swap chain
description is retrieved; specifically, in InitScreenScaleOnce check
SUCCEEDED(globals::d3d::swapChain->GetDesc(&sd)) and, after computing
g_displaySize, g_screenScaleRatio (and after the GetClientRect branch), set
g_scaleInitialized = true there instead of unconditionally at the end—this
ensures failed GetDesc or missing globals::d3d::swapChain will allow retries on
subsequent calls.
- Around line 73-89: Remove the dead static member
OverlayRenderer::screenScaleRatio: delete its declaration and any initialization
in OverlayRenderer.cpp and update any includes/comments referencing it; ensure
all screen-scaling logic continues to use the file-local g_screenScaleRatio used
by InitScreenScaleOnce and UpdateImGuiInput in UI.cpp so no functionality
changes occur.

---

Nitpick comments:
In `@src/Utils/UI.h`:
- Line 52: Add a Doxygen-style doc comment above the declaration of
UpdateImGuiInput in namespace Util describing its purpose, parameters and
behavior (e.g., what hwnd is and how the function updates ImGui input state),
matching the style used by the other public functions in the header so
documentation is consistent; place the comment immediately above the line "void
UpdateImGuiInput(HWND hwnd);" and include `@param` for hwnd and a brief `@brief`
sentence about side effects.

[coderabbitai]

🧹 Nitpick comments (1)

src/Utils/UI.cpp (1)

79-81: Consider gating io.DisplaySize on g_scaleInitialized for consistency.

Currently io.DisplaySize is updated as soon as g_displaySize is non-zero, which can happen while g_screenScaleRatio is still the default 1:1 (if GetDesc succeeds but the first GetClientRect call fails). In that brief transient window, DisplaySize reflects the actual backbuffer but mouse coordinates are unscaled. Using g_scaleInitialized as the guard keeps both in sync:

♻️ Proposed change

-		if (g_displaySize.x > 0 && g_displaySize.y > 0) {
+		if (g_scaleInitialized) {
 			io.DisplaySize = g_displaySize;
 		}

Note: GetClientRect on a valid HWND almost never fails in practice, so this state is transient (≤1 frame) and the current code remains functionally correct.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/Utils/UI.cpp` around lines 79 - 81, Summary: Gate the io.DisplaySize
assignment on g_scaleInitialized to keep display size and mouse scaling
consistent. Fix: in the block that currently sets io.DisplaySize when
g_displaySize.x/y > 0, add a check for g_scaleInitialized (i.e., only assign
io.DisplaySize when g_scaleInitialized && g_displaySize.x > 0 && g_displaySize.y
> 0) so the backbuffer size is applied only after g_screenScaleRatio has been
initialized (the GetDesc/GetClientRect flow involving g_scaleInitialized).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@src/Utils/UI.cpp`:
- Around line 79-81: Summary: Gate the io.DisplaySize assignment on
g_scaleInitialized to keep display size and mouse scaling consistent. Fix: in
the block that currently sets io.DisplaySize when g_displaySize.x/y > 0, add a
check for g_scaleInitialized (i.e., only assign io.DisplaySize when
g_scaleInitialized && g_displaySize.x > 0 && g_displaySize.y > 0) so the
backbuffer size is applied only after g_screenScaleRatio has been initialized
(the GetDesc/GetClientRect flow involving g_scaleInitialized).

[github-actions]

✅ A pre-release build is available for this PR:
Download

[zndxcvbn]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[alandtse]

CI failure isn't an issue. I'll fix it shortly.