Removes the once_cell dependency, instead using std::sync::OnceLock and a
minimal polyfill for std::sync::LazyLock, which may be stabilized soon
(see rust-lang/rust#121377).
This should not require a bump in MSRV, as OnceLock was stabilized in 1.70,
which this crate is using.
Given that `ModifiersChanged` is a window event, it means that clients
may track it for each window individually, thus not sending it between
focus changes may result in modifiers getting desynced on the consumer
side.
Replace the `CustomCursorBuilder` with the `CustomCursorSource` and
perform the loading of the cursor via the
`EventLoop::create_custom_cursor` instead of passing it to the builder
itself.
This follows the `EventLoop::create_window` API.
Creating window when event loop is not running generally doesn't work,
since a bunch of events and sync OS requests can't be processed. This
is also an issue on e.g. Android, since window can't be created outside
event loop easily.
Thus deprecate the window creation when event loop is not running,
as well as other resource creation to running event loop.
Given that all the examples use the bad pattern of creating the window
when event loop is not running and also most example existence is
questionable, since they show single thing and the majority of their
code is window/event loop initialization, they wore merged into
a single example 'window.rs' example that showcases very simple
application using winit.
Fixes#3399.
Mainly fix typos in comments, but also some minor code changes:
* Rename `apply_on_poiner` to `apply_on_pointer`.
* Rename `ImeState::Commited` to `ImeState::Committed`
* Correct `cfg_attr` usage: `wayland_platfrom` -> `wayland_platform`.
While there's a separate event to deliver modifiers for keyboard,
unfortunately, it's not even remotely reflects the modifiers state.
Thus use events along side regular modifier updates to correctly
detect the state. Also, apply the modifiers from the regular
key event by converting their state to xkb modifiers state.
Links: https://github.com/alacritty/alacritty/issues/7549Closes: #3388
We could fail to setup property watcher and fail to start, thus
don't require XSETTINGS to work.
Fixes: df8805c0 (On X11, reload DPI on _XSETTINGS_SETTINGS)
This also fixes the deadlock when such reload may happen.
Fixes: #3383
Signed-off-by: John Nunley <dev@notgull.net>
Signed-off-by: Kirill Chibisov <contact@kchibisov.com>
This let statement swapped the two names, resulting in incorrect
behavior since commit d7ec899d. That commit did not actually introduce
the swap, but the previous code swapped it again before setting the
WM_CLASS property, so no issue was ever observed.
It also brings the documentation in line with the implementation since the
parent commit, and with the ICCCM standard, which states the following
about the WM_CLASS property [1]:
The two strings, respectively, are:
* A string that names the particular instance of the application [...]
* A string that names the general class of applications [...]
[1] https://www.x.org/releases/current/doc/xorg-docs/icccm/icccm.html#WM_CLASS_Property
xdotool will update modifiers before Xkb will actually send event
updating them, thus the modifiers will be updating even before the
actual update, which is unfortunate.
Links: https://github.com/alacritty/alacritty/issues/7502
This appears to be the solution for the elusive #3335 issue. Previously,
in the Xlib backend, we used the "XRRQueryVersion" function to query for
the Xrandr version, and used that to determine whether we should use the
"GetScreenResources" call or the "GetScreenResourcesCurrent" call.
However, we passed the version "0, 0" into "XRRQueryVersion".
Previously with Xlib this wasn't a problem, as Xlib ignores the version
you pass in and substitutes it with the version of RandR it expects.
https://gitlab.freedesktop.org/xorg/lib/libxrandr/-/blob/master/src/Xrandr.c?ref_type=heads#L386-387
The way that "XRRQueryVersion" is implemented on the server end, it
compares the version passed into the request with the version supported
by the server. If the server's version is greater than the client
version, it just returns the client version. If the client's version is
greater, it passes the server's version. Since we were passing in "0, 0"
this means that the server returned "0, 0".
https://gitlab.freedesktop.org/xorg/xserver/-/blob/master/randr/rrdispatch.c?ref_type=heads#L50-59
To determine whether we use "GetScreenResources" or
"GetScreenResourcesCurrent", we compare the version returned by the
server against "1, 3". Since we got "0, 0"- a version of XRandR so old
it doesn't even exist- we use "GetScreenResources".
The problem manifests in that "GetScreenResources" can take several
seconds to query the screen state based on the current hardware
configuration. On the other hand, "GetScreenResourcesCurrent" is fast;
it uses the server's hardware cache if it is available.
This problem is visible in XTrace. On the latest `master`:
```
000:<:00c2: 12: RANDR-Request(140,0): QueryVersion major-version=0 minor-version=0
000:>:00c2:32: Reply to QueryVersion: major-version=0 minor-version=0
000:<:00c3: 8: RANDR-Request(140,8): GetScreenResources window=0x0000076e
000:>:00c3:1600: Reply to GetScreenResources:
```
On the `v0.28.0` tag:
```
000:<:0019: 12: RANDR-Request(140,0): QueryVersion major-version=1 minor-version=6
000:>:0019:32: Reply to QueryVersion: major-version=1 minor-version=6
...later
000:<:002d: 8: RANDR-Request(140,25): GetScreenResourcesCurrent window=0x0000076e
000:>:002d:1600: Reply to GetScreenResourcesCurrent
```
This commit fixes this issue by requesting "1, 3" instead. This returns
the version we expect, where we can now use "GetScreenResourcesCurrent"
properly.
Fixes#3335
Signed-off-by: John Nunley <dev@notgull.net>
This code confused me. I tried to understand it. I tried to simplify it
while keeping the functional style. But in the end, this just seems too
complicated for its own good. Just doing the exact same thing with a
match statement and the question mark operator makes it sooo much more
obvious what is happening.
Signed-off-by: Uli Schlachter <psychon@znc.in>
There seems to be many PRs relating to this issue, but they don't include all
platforms and for some reason lost steam. This PR again tries to make this
feature happen, and does it for all desktop platforms (x11, wayland, macos,
windows, web).
I think the best user of this feature and the reason I'm doing this is Bevy and
game engines in general. There non laggy hardware cursors with custom images are
very important. Game devs also like their PNGs so supporting platform native
cursor files is not that important, but I guess could be added too.
Co-authored-by: daxpedda <daxpedda@gmail.com>
Co-authored-by: Mads Marquart <mads@marquart.dk>
Co-authored-by: Kirill Chibisov <contact@kchibisov.com>
The only breaking change is that x11rb no longer reports an error when
querying the WmSizeHints of a window that does not have this property
set. For this reason, the return type of WmSizeHintsCookie::Reply()
changed from Result<WmSizeHints, SomeError> to
Result<Option<WmSizeHints>, SomeError>.
In update_normal_hints(), previously a cryptic error would be reported
to the caller. Instead, this now uses unwrap_or_default() to get a
WmSizeHints. All fields of WmSizeHints are Options, so this produces an
empty object.
resize_increments() queries a value from the window and returns an
Option. Previously, the error for "missing property" was turned into
None via .ok(). This commit adds a call to flatten() to also turn
"property not set" into None.
Finally, request_user_attention() queries a window's WmHints property
and updates one field of it. The code already uses unwrap_or_default()
to deal with missing properties, so just a call to flatten() is needed
to merge "missing property" and "error while querying" into one.
Other changes in x11rb do not seem to affect this crate.
x11rb's MSRV increased from 1.56 to 1.63, which is still below the MSRV
of this crate, which is 1.65.
Signed-off-by: Uli Schlachter <psychon@znc.in>
Closes#3245
notgull forgot to properly interpret float data from the X server,
making him tonight's biggest loser.
Signed-off-by: John Nunley <dev@notgull.net>
Keep the user provided size in the original values and convert only
when we're getting a `configure` event. On some compositors will
have a scale available, so it'll work, however with some we'll
still have old 'pick 1` as default.
Also configure_bounds when compositor tells the user to pick the size,
that will ensure that initial `with_inner_size` won't grow beyond the
working area.
Fixes#3187.
The cursor hittest was not reloaded on window size changes, only
when `Window::request_inner_size` was called leading to regions
of the window being not clickable.
Also, don't try to apply hittest logic when user never requested a
hittest.
Links: https://github.com/alacritty/alacritty/pull/7220
Split `Key` into clear categories, like `Named`, `Dead`, Character`, `Unidentified`
removing the `#[non_exhaustive]` from the `Key` itself.
Similar action was done for the `KeyCode`.
Fixes: #2995
Co-authored-by: Kirill Chibisov <contact@kchibisov.com>
Window builder is always accessed by winit on the thread event loop
is on, thus it's safe to mark the data it gets as `Send + Sync`.
Each unsafe object is marked individually as `Send + Sync` instead
of just implementing `Send` and `Sync` for the whole builder.
