[v3, macOS] Screen.Bounds.X/Y on non-primary screens are off by the screen's scale factor (mixed-DPI)

Description

In pkg/application/screen_darwin.go, cScreenToScreen multiplies screen.x and screen.y by the screen's own backingScaleFactor before assigning them to Bounds.X / Bounds.Y:

sf := float64(screen.scaleFactor)
toPhysical := func(points C.int) int { return int(float64(points) * sf) }

return &Screen{
    X: toPhysical(screen.x),
    Y: toPhysical(screen.y),
    Bounds: Rect{
        X: toPhysical(screen.x),
        Y: toPhysical(screen.y),
        ...
    },
    ...
}

The intent (per the code comment) is a round-trip workaround for applyDPIScaling in screenmanager.go, which divides Bounds.Width / Bounds.Height by the scale factor. But Bounds.X / Bounds.Y are not divided again in applyDPIScaling — they're left in points × ownScaleFactor space, while Bounds.Width / Bounds.Height end up in logical points.

After #5304 + the GetScreens normalization, Position() / SetPosition() are in logical points across screens. So on mixed-DPI setups (e.g., a Retina secondary above a 1x primary), SetPosition(s.Bounds.X, s.Bounds.Y) for a non-primary screen lands at the wrong location by a factor of that screen's ScaleFactor.

Steps to Reproduce

1. Primary monitor: 1x (non-Retina).
2. Secondary monitor: 2x Retina, positioned above the primary in macOS Displays settings.
3. Read secondary.Bounds.Y from GetScreens().
4. Call SetPosition(secondary.Bounds.X, secondary.Bounds.Y).
5. Window does not land at the top-left of the secondary; it's off by secondary.ScaleFactor.

The same ScaleFactor doubling exists today but only becomes user-visible once Position/SetPosition are normalized to logical points (which is the goal of #5117 / #5304).

Notes

• The flofreud setup in #5117 (Retina primary + 1x external above) is not affected, because the non-primary screen has ScaleFactor=1 and the multiplication is a no-op.
• This is a pre-existing macOS-only Bounds unit bug — exposed (not introduced) by the GetScreens normalization landing alongside #5304.
• The fix is likely: drop the * sf multiplication on X/Y in cScreenToScreen (keep it for PhysicalBounds.X/Y), and verify nothing in screenmanager.go (e.g., move() placement adjustment, getScreenAlignment) regresses on the cross-DPI tests.

Environment

• Wails v3 (pkg/application/screen_darwin.go, pkg/application/screenmanager.go)
• macOS, mixed-DPI multi-monitor

Related

• #5117 — coordinate-system audit
• #5304 — absolute coordinate normalization PR (where this case became user-visible)

Start with cScreenToScreen in pkg/application/screen_darwin.go, then inspect applyDPIScaling, move(), and getScreenAlignment in pkg/application/screenmanager.go. Run the cross-DPI tests and verify that Bounds coordinates remain in logical points, PhysicalBounds coordinates retain physical scaling, and SetPosition can round-trip a non-primary screen position.