Converting positions and rectangles in screen physical pixels to screen DIPs (device-independent pixels) in Chromium -- and vice versa -- is handled differently on Windows compared to other platforms. This difference is due to Windows' historical challenges with handling displays of DPIs higher than 96 and the need to deal with multiple DPI displays.
On Windows, the process involves dealing with monitor physical coordinates and mapping them into Chrome's DIP (Device Independent Pixels) coordinates. The key points in this conversion process are as follows:
Monitor Tree: DisplayInfosToScreenWinDisplays reasons over monitors as a tree using the primary monitor as the root. All monitors touching this root are considered children.
Connected Components: This approach presumes that all monitors are connected components. Windows restricts the layout of monitors to connected components, except when DPI virtualization is happening. In the case of DPI virtualization, scaling is relative to the (0, 0) point.
Overlap Handling: There can be cases where a scaled display may have insufficient room to lay out its children. In these cases, a DIP point could map to multiple screen points due to overlap. The first discovered screen point will take precedence.
On other platforms, such as macOS and Linux, the display information received from the OS is already adjusted for display scaling. This makes it easier to convert between screen coordinates and screen DIPs by simply multiplying or dividing by the current screen's display scale factor.
Windows:
ScreenWin class (ui/display/win/screen_win.cc). This class takes into account the monitor tree, connected components, and overlap handling when converting coordinates.display::win::GetScreenWin()->DIPToScreenRect(GetParentHWND(), dip_bounds) will convert a screen DIP rect to a screen physical one.Other Platforms: