feat: stable ids

a11y: Don't unconditionally pull winit (pop-os#43)

fix: quad rendering including border only inside of the bounds

fix: better slider drawing (it allows just the border part of the handle quad outside of the layout bouds, which isn't great, but is ok for our purposes due to being transparent)

cleanup: fix & format

fix: use iced_core::Font

cleanup

fix: allow leaving out winit & iced-sctk

fix: settings

fix: slider draw improvements

fix: websocket example

fix: modal example

fix: scrollable example

fix: toast example

fix: avoid panicking in iced_sctk with lazy widgets in auto-size surfaces

fix: todos panic

fix: only diff auto-sized surfaces in iced_sctk build_user_interface & improve sctk examples

wip (iced-sctk): window resize with icons

feat (iced-sctk): support for setting cursor

refactor: default decorations to client

fix: set window geometry after receiving configure

fix: size limits with no max bound must be cut off

fix: send size update when autosized surface resizes

fix: use ceil size for positioner

cleanup: remove dbg statement

fix: remove a destroyed surface from compositor surfaces

fix errors after rebase and wip scaling support

fix: handling of scale factor in set_logical_size

fix (sctk_drag example): add .into for border radius

fix: fractional scaling

sctk: Fire RedrawRequests

wip: animations via frame event

fix / refactor iced-sctk redraw & frame event handling

cleanup: note about frame request in iced-sctk

fix: send resize when necessary for layer surface and popups too

fix: always request redraw for a new surface

fix: scaling and autosize surface improvements

refactor: sctk_lazy keyboard interactivity

feat(sctk): configurable natural_scroll property

feat: send state and capabilities events when there are changes

fix: redraw when an update is needed and clean up the logic

Update sctk to latest commit

Fix compilation of sctk drag example

fix(sctk): update interface before checking if it has a redraw request

refactor: after autosize surface resize wait to redraw until the resize has been applied

refactor: better handling of autosize surfaces

chore: update sctk

chore: update sctk

fixes sctk_drag example

fix: default to ControlFlow::Wait for applications with no surface

this seems to help CPU usage for app library and launcher

default to 250ms timeout in the event loop

Update sctk

sctk: Implement xdg-activation support

fix: don't require Flags to be clone for settings on wayland

chore: error if neither winit or wayland feature is set

chore: Allow compiling without windowing system (pop-os#65)

fix(iced-sctk): handle exit_on_close_request

fix: make sure that each widget operation operates on every interface

This should be ok even for widget actions like focus next because there can only ever be a single focused widget

cargo fmt

cleanup: dbg statement

fix(iced-sctk): replace panic with handling for remaining enum variants

refactor: use iced clipboard for interacting with the selection

refactor: allow passing an activation token when creating a window

sctk: Add support for `ext-session-lock` protocol

fix(sctk): build and use tree for layout of autosize surfaces

Update winit to latest commit used by upstream iced

fix(sctk): send key characters

fix(sctk): check if key is a named key first

refactor(sctk): keep compositor surface in state

…he window::screenshot command.

Co-Authored-By: Austin M. Reppert <[email protected]>

By default, this is the same as the text color for best visibility.

Part of this is a refactor of the ID

cleanup: clippy and fmt

fix: test workflow

fix: add note in CHANGELOG

fix: clippy

…able

A `dnd_listener` widget shouldn't handle a DnD event when the dnd drag
isn't within the widget's bounds. So add a few more checks for this.

Enter/leave events generated by `DndOfferEvent::Motion` also don't
behave as one might expect, since the enter may occur before the leave
depending on the order it calls `on_event` on the widget. Not sure how
to address that, but cosmic-workspaces can just ignore the leave events
for now.

Otherwise, this seems to be working fine, after these changes.

This can be tested with something like
`VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/nvidia_icd.json cargo run -p tour
--features iced/wgpu`.

On Nvidia I'm seeing a flood of `Suboptimal present of frame` warnings.
So some improvement may still be needed here. But if it doesn't regress
behavior on other hardware, that seems like an improvement over
freezing.

This adds a widget that attaches an shm or dma buffer to a subsurface,
scaled with `wp_viewporter`.

By exposing this as a widget, rather than as a type of window, it can be
positioned and scaled like any other iced widget. It provides an API
that's similar to an iced image.

The initial version of this just took a `wl_buffer`. But this makes
buffer re-use problematic. In particular, the docs for `wl_surface::attach`
note that `wl_buffer::release` events become unreliable if a buffer is
attached to multiple surfaces. And indicates that a client should create
multiple `wl_buffer` instances, or use `wp_linux_buffer_release`.

So we store information about the buffer, and create `wl_buffer`s as
needed. `SubsurfaceBuffer::new` also returns a future that's signaled
when all references are destroyed, both `wl_buffer`s and any instance of
the `SubsurfaceBuffer` that might still be used in the `view`.

So this seems like the best solution for now, within the
model-view-update architecture.

This has two examples: `sctk_subsurface`, showing a single-color shm
buffer, and `sctk_subsurface_gst`, which plays an h264 video to a
subsurface with vaapi decoding.

This was particularly visible on Redox where there is no vsync, but also 
causes unnecessary redraws on Linux

Winit currently supports this only on Windows and Wayland.

This requests that a context menu is shown at the cursor position,
like the menu normally triggered by right clicking the title bar. This
is important for implementing client side decorations with Iced widgets.

…is closed

Autosized surfaces perform the layout step to get the size and then again when building the interface, but sometimes the calculated size is not enough space when used as a bound, so we need to add a tiny amount to the calculated size. This also makes the event loop timeout duration configurable. Viewport physical size is calculated directly from the logical size now as well in iced-sctk to avoid inconsistencies that resulted from recalculating the logical size after using it to calculate the physical size.

Creating a new `wl_buffer` each frame seems to perform poorly. We can
instead keep a cache of `wl_buffer`s we have created from a
`BufferSource`.

Similar to `waylandsink`. Allows us to avoid creating a buffer source
(and ultimately `wl_buffer`) for every buffer swap.

If the main surface is occluded completely by opaque subsurfaces, it may
not receive `frame` events. So we need to request frame events for all
subsurfaces as well.

Additionally, with "synchronized" subsurfaces, we need to `commit` the
parent surface for subsurface changes to take effect.

Fixes issues with subsurfaces updating slowly, or only when mouse moved
under some circumstances.

Whether or not this works seems to depend on driver, or gstreamer
version...

Useful for testing pointer input to subsurfaces.

when broadcasting events for no specific surface, it should be done after update so that the runtime subscription is current

(cherry picked from commit f39a5fd)

A null `region` represents an infinite region (the default). To set an
empty region, we need to create a `wl_region`.