Signed-off-by: Marc <marc.leroy@samantree.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

The Windows Pointer API
(https://learn.microsoft.com/en-us/windows/win32/api/_inputmsg/) was used
to allow the detection of events when interacting with touch screens. This
also opens the gates for supporting other types of input devices (e.g. pens
and touchpads). Mouse events are now part of the pointer events (primary
events trigger WM_POINTER<DOWN|UP>, whereas secondary ones vanilla
WM_POINTERUPDATE, and cancellations WM_POINTERCAPTURECHANGED).
A fourth and fifth button (usually found in modern mice) has also been added
for completeness, though their integration in other OS-es shall be the
objective of future patches.

Signed-off-by: Marc <marc.leroy@samantree.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Currently build fails as go.mod uses go 1.23.8 which doesn't have
strings.SplitSeq.  Note: strings.SplitSeq was introduced in go 1.24.

Signed-off-by: Miles Alan <m@milesalan.com>

Fixes: https://todo.sr.ht/~eliasnaur/gio/654
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

The normal vector size and direction depend on the input point and the sign of the unit value provided.

Fixes: https://todo.sr.ht/~eliasnaur/gio/576
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

There are two max window size settings on macOS, `contentMaxSize` and
`maxFullScreenContentSize`. Set the latter to avoid a window being
resized larger than its maximum in full screen mode.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

According to @kkeybbs, pressing the maximize button on Windows only
resizes the window up to its maximum bounds. That means we can leave the
button available, and only hide it when the window has a fixed size.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Fixes: https://todo.sr.ht/~eliasnaur/gio/656
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Cancel events affect the entire gesture and as such all active pointers.

References: https://todo.sr.ht/~eliasnaur/gio/657
Signed-off-by: Elias Naur <mail@eliasnaur.com>

This is a breaking change, users that need the old behavior can implement it using the existing API.

Fixes: https://todo.sr.ht/~eliasnaur/gio/511
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Avoids unnecessary eglMakeCurrent and thread locking calls to enhance performance.

Fixes: https://todo.sr.ht/~eliasnaur/gio/658
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Reduces ambiguity by introducing AffineId() for representing identity transformation matrices.

References: https://todo.sr.ht/~eliasnaur/gio/655
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

References: https://todo.sr.ht/~eliasnaur/gio/655
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

References: https://todo.sr.ht/~eliasnaur/gio/655
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

After moving the pen, the next action should update the bounds for the start position.

References: https://todo.sr.ht/~eliasnaur/gio/451
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Fixes: https://todo.sr.ht/~eliasnaur/gio/534
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Fixes the edge case where a zero point would be normalized to NaN.

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

This fixes the cases where the unit length matches the hypotenuse.

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>

Fixes the 1px overlap of curve quads. Without this patch the rendered quads
were skewed vertically and were 2 pixels shorter in height. The NorthWest
pixels were moved to the SouthWest instead of NorthWest and the SouthEast
pixels were moved to the NorthEast instead of SouthEast.

References: https://todo.sr.ht/~eliasnaur/gio/339
Signed-off-by: Walter Werner SCHNEIDER <contact@schnwalter.eu>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Using cgo.Handle allows us to pass a reference to a Go function through
the GCD API for running main thread code, saving a goroutine and a channel.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Inspired by the discussion at golang.org/issue/70089, this change makes
our particular NSApplicationDelegate implementation optional.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Our go.mod says 1.24, so a go1.18 conditional is redundant.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Floating windows are rendered above all other non-floating windows.

Apple Documentation: https://developer.apple.com/documentation/appkit/nswindow/level-swift.struct

Signed-off-by: Thomas Banks <thomas@tombanks.me>

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Signed-off-by: inkeliz <inkeliz@inkeliz.com>

References: https://lists.sr.ht/~eliasnaur/gio/%3CCAMAFT9Uyh_JWrkQQt+AmekJWFBqhZPsP_3ZxC1fUNB+=VGGorw@mail.gmail.com%3E
Signed-off-by: CoyAce <akeycoy@gmail.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

UIScene notifications are the correct way to track window focus on
iOS 13.0+, because the Key Window API is scene-level on iOS 13.0+,
but we need to maintain support for iOS 12 and earlier.

Implementation strategy:
- iOS 13.0+: Use UIScene notifications (DidActivate/WillDeactivate)
- iOS 12 and earlier: Keep existing UIWindow notifications as fallback
- Add runtime version checks to select the appropriate API

Key changes in os_ios.m:
- Add @available(iOS 13.0, *) checks
- Register both notification types with conditional logic
- Ensure proper cleanup of observers when moving between windows

See: https://developer.apple.com/documentation/uikit/uiscene?language=objc

Fixes: https://lists.sr.ht/~eliasnaur/gio/%3CCAMAFT9Uyh_JWrkQQt+AmekJWFBqhZPsP_3ZxC1fUNB+=VGGorw@mail.gmail.com%3E
Signed-off-by: CoyAce <akeycoy@gmail.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

…stroyed

Config.Focused represents window interaction focus, which should only
change when the window gains or loses user interaction capability.

The focus state is already correctly handled in:
- onResume → focused = true  (window gains interaction focus)
- onPause  → focused = false (window loses interaction focus)

Therefore, sending additional ConfigEvent in onStop and onSurfaceDestroyed
is redundant because:
- onStop is always preceded by onPause
- onSurfaceDestroyed is a low-level surface event unrelated to focus

This cleanup aligns the Android implementation with the correct semantic
that Config.Focused = window interaction focus, not view focus or
surface state.

Signed-off-by: CoyAce <akeycoy@gmail.com>

Previously, the Shortcut key was hardcoded as ModCtrl. That
patches tries to identify the OS and change the key.

Fixes: https://todo.sr.ht/~eliasnaur/gio/624
Signed-off-by: Lucas Rodrigues <inkeliz@inkeliz.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Fix several issues in the IME implementation

- Typing after an editor regains focus will not add text at the begging of the Editor.
- On Safari iOS, selecting a keyboard suggestion no longer duplicates the text.
- Replacing selected text now occurs at the correct position.

This patch change how "input" event is handled, instead of consider all events the
same.

Signed-off-by: Lucas Rodrigues <inkeliz@inkeliz.com>

Previously, using SVG fonts will cause Gio to render invisible
"characters".

Now, some fonts (like Noto Sans Emoji) will be rendered
based on "Outline". Gio don't support SVG fonts, but now
it will show the outline ("black-and-white") alternative.

Signed-off-by: Lucas Rodrigues <inkeliz@inkeliz.com>

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Click and Hover both stored the first PointerID they observed in
their internal pid field and only updated it when not currently
hovered/entered. Once the gesture became hovered, any later event
under a different PointerID was effectively ignored: Click.Press
fell through 'c.pid != e.PointerID' and was silently dropped, and
Hover could never reset entered when the matching Leave arrived
under a new ID.

The Windows backend enables EnableMouseInPointer
(app/os_windows.go), under which Windows reassigns the same
physical mouse's PointerID across focus changes, window
leave/re-enter, and similar events. Once a widget had been hovered,
every subsequent press on it failed to register, including
widget.Editor's internal clicker that positions the caret on press.
Multi-line editors silently refused to move the caret on click
after the window had received any focus event.

Always take the latest PointerID on Hover.Enter and Click.Press.
The Press/Release handshake still works because Press now records
the press's own PointerID and Release continues to gate on
'c.pid != e.PointerID' so an unrelated pointer's release can't end
the press tracking.

Signed-off-by: Eugene <eugenebosyakov@gmail.com>

Fixes: https://todo.sr.ht/~eliasnaur/gio/681
Signed-off-by: Chris Waldon <christopher.waldon.dev@gmail.com>

Scroll events arrived at pointerQueue.Push and went through pointerOf
+ deliverEnterLeaveEvents + deliverEvent like Move/Press/Release. The
side effect: every scroll created or updated a state.pointers entry,
populated p.entered with whatever handlers sat under the wheel
position, and overwrote state.cursor based on hit-test at the scroll
position.

When the platform layer reports scroll with a different PointerID
than mouse-move events for the same physical mouse — which the
Windows backend does (scrollEvent omits PointerID, defaulting to 0,
while pointerUpdate forwards Windows' assigned ID) — the scroll
spawns a phantom state.pointers entry. Subsequent moves go to the
mouse's "real" entry, so the phantom never receives Leave events,
its entered set never empties, and the cleanup at the end of Push
keeps it alive. pointerQueue.Frame then runs hit-test for it every
frame at the user's last scroll position, threading state.cursor
through it after the live pointer's resolution and clobbering it
with whatever's under the scroll position.

The wheel is positional but isn't a pointer. Treating it as one is
the bug. Hit-test inline at the scroll position to find delivery
targets, dispatch via deliverEvent (which already handles filter
matching, scroll axis clamping, and area-local position), and
return without creating or updating a state.pointers entry.

Add a router-level test that fails without the fix: a Move sets the
cursor over a CursorPointer region, a subsequent Scroll over a
CursorText region, and the test asserts the cursor is still
CursorPointer. Pre-fix the scroll's deliverEnterLeaveEvents
overwrites state.cursor with CursorText.

Signed-off-by: Eugene <eugenebosyakov@gmail.com>

typesetting introduced a cache field that needs to be
properly initialized. Use constructor to avoid the issue.

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

This way their cache can be shared.

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

Signed-off-by: Egon Elbre <egonelbre@gmail.com>

loadEGL used sync.Once incorrectly: the error returned by
loadDLLs was assigned to a local variable inside loadEGL,
so on the second call Do would not run and the
function would return nil even though the DLLs were never
loaded. This caused a nil pointer dereference when callers
proceeded to use _eglGetDisplay and other uninitialized
function pointers.

Fix by replacing the sync.Once with sync.OnceValue, which
correctly caches the return value of loadDLLs across all
calls.

Signed-off-by: Kevin Yuan <farproc@gmail.com>

DefWindowProc handles WM_WINDOWPOSCHANGED by sending WM_SIZE and WM_MOVE
messages, which would lead us to handle resizes twice.

Per MSDN, the WM_SIZE handler is made redundant by handling
WM_WINDOWPOSCHANGED:
https://learn.microsoft.com/en-us/windows/win32/winmsg/wm-windowposchanged

Signed-off-by: Kevin Yuan <farproc@gmail.com>
Signed-off-by: Elias Naur <mail@eliasnaur.com>

Decorations buttons already use widget.Clickable and draw press ink, but they don't draw a hover/focus background. This makes custom window decorations appear unresponsive when the pointer is over minimize, maximize, or close buttons.

Register each button's system action over its clickable area and draw the same hovered background used by other material buttons. This also lets platforms query the button action from the material decorations hit area.

Signed-off-by: qiannian <qianniancn@gmail.com>

Keep custom move areas mapped to HTCAPTION even when the window is
maximized so custom title bars preserve the standard Windows behavior
where double-click restores the window.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Allow platform backends to send pointer.Leave directly.
The router delivers it to entered handlers so hover state is cleared normally.

Signed-off-by: Elias Naur <mail@eliasnaur.com>

Mark WM_IME_STARTCOMPOSITION as handled after positioning the IME
composition and candidate windows.

Passing the message on to DefWindowProc can let Windows create its default
composition window, which shows the first composing character below Gio's
caret.

Fixes: https://todo.sr.ht/~eliasnaur/gio/687
Signed-off-by: qiannian <qianniancn@gmail.com>

Track IME composition range updates as key.CompositionEvent and route them to
the focused editor. Draw a thin underline under the current composition range
using the editor text material.

Keep the composition range in imeState so it is reset with the rest of the
editor input method state, and normalize it in input.Router before forwarding
CompositionEvent.

Signed-off-by: qiannian <qianniancn@gmail.com>