* Add support for generating dummy DeviceIDs and WindowIDs
* Fix linux
* Improve docs and move dummy to unsafe
* Strengthen guarantees a bit
* Add backticks to CHANGELOG.md
Co-Authored-By: Xaeroxe <xaeroxe@amethyst-engine.org>
This commit restricts an Xfwm4-specific DPI-preserving hack to Xfwm4
only. The hack saves and restores the DPI-adjusted size until the actual
size matches. On tiling WMs like i3 this fails, since the size is
constrained by the layout. This in turn causes a never-ending
XResizeWindow vs. XConfigureWindow fight between the WM and the client,
making the WM, winit client, and Xorg consume all CPU cycles available.
Fixes#691
Dropping a window before running the `EventsLoop` results in events
still being queued when `XDestroyWindow` is called, so events like
`XI_Enter` (the culprit in this case) will still be processed.
Simply checking that the window still exists before calling
`query_pointer` was enough to solve the problem.
* x11: compute resize logical size with new dpi
Whenever a dpi change occurs, trigger a Resized event as well with the
new logical size. Given X11 primarily deals in physical pixels, a change
in DPI also changes the logical size (as the physical size remains
fixed).
* Doc tweaks
* X11: General cleanup
This is almost entirely internal changes, and as usual, doesn't actually
fix any problems people have complained about.
- `XSetInputFocus` can't be called before the window is visible. This
was previously handled by looping (with a sleep) and querying for the
window's state until it was visible. Now we use `XIfEvent`, which blocks
until we receive `VisibilityNotify`. Note that this can't be replaced
with an `XSync` (I tried).
- We now call `XSync` at the end of window creation and check for
errors, assuring that broken windows are never returned. When creating
invisible windows, this is the only time the output buffer is flushed
during the entire window creation process (AFAIK). For visible windows,
`XIfEvent` will generally flush, but window creation has overall been
reduced to the minimum number of flushes.
- `check_errors().expect()` has been a common pattern throughout the
backend, but it seems that people (myself included) didn't make a
distinction between using it after synchronous requests and asynchronous
requests. Now we only use it after async requests if we flush first,
though this still isn't correct (since the request likely hasn't been
processed yet). The only real solution (besides forcing a sync *every
time*) is to handle asynchronous errors *asynchronously*. For future
work, I plan on adding logging, though I don't plan on actually
*handling* those errors; that's more of something to hope for in the
hypothetical async/await XCB paradise.
- We now flush whenever it makes sense to. `util::Flusher` was added to
force contributors to be aware of the output buffer.
- `Window::get_position`, `Window::get_inner_position`,
`Window::get_inner_size`, and `Window::get_outer_size` previously all
required *several* round-trips. On my machine, it took an average of
around 80µs. They've now been reduced to one round-trip each, which
reduces my measurement to 16µs. This was accomplished simply by caching
the frame extents, which are expensive to calculate (due to various
queries and heuristics), but change infrequently and predictably. I
still recommend that application developers use these methods sparingly
and generally prefer storing the values from `Resized`/`Moved`, as
that's zero overhead.
- The above change enabled me to change the `Moved` event to supply
window positions, rather than client area positions. Additionally, we no
longer generate `Moved` for real (as in, not synthetic)
`ConfigureNotify` events. Real `ConfigureNotify` events contain
positions relative to the parent window, which are typically constant
and useless. Since that position would be completely different from the
root-relative positions supplied by synthetic `ConfigureNotify` events
(which are the vast majority of them), that meant real `ConfigureNotify`
events would *always* be detected as the position having changed, so the
resultant `Moved` was multiple levels of misleading. In practice, this
meant a garbage `Moved` would be sent every time the window was resized;
now a resize has to actually change the window's position to be
accompanied by `Moved`.
- Every time we processed an `XI_Enter` event, we would leak 4 bytes via
`util::query_pointer` (`XIQueryPointer`). `XIButtonState` contains a
dynamically-allocated mask field which we weren't freeing. As this event
occurs with fairly high frequency, long-running applications could
easily accumulate substantial leaks. `util::PointerState::drop` now
takes care of this.
- The `util` module has been split up into several sub-modules, as it
was getting rather lengthy. This accounts for a significant part of this
diff, unfortunately.
- Atoms are now cached. Xlib caches them too, so `XInternAtom` wouldn't
typically be a round-trip anyway, but the added complexity is
negligible.
- Switched from `std::sync::Mutex` to `parking_lot::Mutex` (within this
backend). There appears to be no downside to this, but if anyone finds
one, this would be easy to revert.
- The WM name and supported hints are now global to the application, and
are updated upon `ReparentNotify`, which should detect when the WM was
replaced (assuming a reparenting WM was involved, that is). Previously,
these values were per-window and would never update, meaning replacing
the WM could potentially lead to (admittedly very minor) problems.
- The result of `Window2::create_empty_cursor` will now only be used if
it actually succeeds.
- `Window2::load_cursor` no longer re-allocates the cursor name.
- `util::lookup_utf8` previously allocated a 16-byte buffer on the heap.
Now it allocates a 1024-byte buffer on the stack, and falls back to
dynamic allocation if the buffer is too small. This base buffer size is
admittedly gratuitous, but less so if you're using IME.
- `with_c_str` was finally removed.
- Added `util::Format` enum to help prevent goofs when dealing with
format arguments.
- `util::get_property`, something I added way back in my first winit PR,
only calculated offsets correctly for `util::Format::Char`. This was
concealed by the accomodating buffer size, as it would be very rare for
the offset to be needed; however, testing with a buffer size of 1,
`util::Format::Long` would read from the same offset multiple times, and
`util::Format::Short` would miss data. This function now works correctly
for all formats, relying on the simple fact that the offset increases by
the buffer size on each iteration. We also account for the extra byte
that `XGetWindowProperty` allocates at the end of the buffer, and copy
data from the buffer instead of moving it and taking ownership of the
pointer.
- Drag and drop now reliably works in release mode. This is presumably
related to the `util::get_property` changes.
- `util::change_property` now exists, which should make it easier to add
features in the future.
- The `EventsLoop` device map is no longer in a mutex.
- `XConnection` now implements `Debug`.
- Valgrind no longer complains about anything related to winit (with
either the system allocator or jemalloc, though "not having valgrind
complain about jemalloc" isn't something to strive for).
* X11: Add better diagnostics when initialization fails
* X11: Handle XIQueryDevice failure
* X11: Use correct types in error handler
* x11: Always receive Awakened event in run_forever
Do not reset the pending_wakeup boolean at the start of run_forever so
that each call to EventsLoopProxy::wakeup results in an Awakened event.
Fixes#462
* Update CHANGELOG.md
* Replace Closed event with CloseRequested and Destroyed
Implements #434
The existing Closed event had ambiguous meaning, both in name and in
cross-platform behavior. Closed is now split into two more precise events:
* CloseRequested - the window has been requested to close, most commonly by
having clicked the window's close button. Whether or not you respond by
closing the window is up to you.
* Destroyed - the window has been destroyed, and can no longer be safely
used.
Most notably, now you can reliably implement classic patterns like
prompting the user to save their work before closing, and have the
opportunity to perform any necessary cleanup.
Migrating to the new API is straightforward. In most cases, you can simply
replace all existing usages of Closed with CloseRequested. For more
information, see the example programs, particularly handling_close and
multiwindow.
iOS applications must replace all usages of Closed with Destroyed, and
require no other changes.
* x11: Windows are Sync again
Fixes#472
* Add test ensuring that Window is Sync
Window must be Sync for Vulkano's Arc<FramebufferAbstract> to be usable.
Fixes#195Fixes#277Fixes#455
* Read `XMODIFIERS` explicitly/directly instead of calling `XSetLocaleModifiers` with an
empty string. This is useful for debugging purposes, and more clear to read and handle.
* Fallback to local input method if the one specified in `XMODIFIERS` is later closed on the
server end (i.e. if ibus/fcitx is terminated). Previously, that would cause the event loop
to freeze and usually also segfault.
* If using the fallback input method, respond to the `XMODIFIERS` input method later
becoming available. This means that the input method restarting is handled, and that even if
the program was started while ibus/fcitx/etc. was unavailable, it will start using it as
soon as it becomes available.
* Only one input method is opened for the whole event loop, with each window having its own
input context.
* IME works completely out of the box now, no longer requiring application developers to
call `setlocale` or `XSetLocaleModifiers`.
* Detailed error messages are provided if no input method could be opened. However, no
information is provided to the user if their intended `XMODIFIERS` input method failed to
open but the fallbacks (which will ostensibly always succeed) succeeded; in my opinion, this
is something that is best filled by adding a logging feature to winit.
XIM isn't thread-safe at all. Any call made to it from another thread will result in the
event loop freezing (this is why the old implementation of Drop for Window had that
problem).
XIM is now confined to one thread, and the existing API is maintained using channels. In
testing this with Alacritty, I initially thought the occasional slight lag on updating the
spot location was due to this change, but it's present without it as well.
* Try XOpenIM with different locale modifiers
Implements the solution suggested in
https://github.com/tomaka/winit/issues/277#issuecomment-337751136.
* Use empty XSetLocaleModifiers beforehand
Also, for modifiers, convert from length-based UTF-8 strings to
null-terminated bytestrings.
* Add CHANGELOG entry and comments
Fixes#79#414
This changes the implementation of Drop for Window to send a WM_DELETE_WINDOW ClientMessage,
offloading all the cleanup and window destruction to the event loop. Unsurprisingly, this
entails that the event loop now handles WM_DELETE_WINDOW using the behavior that was
previously contained in Window's Drop implementation, along with destroying the Window.
Not only does this mean that dropped windows are closed, but also that clicking the × button
on the window actually closes it now.
The previous implemention of Drop was also broken, as the event loop would be (seemingly
permenanently) frozen after its invocation. That was caused specifically by the mutex
locking, and is no longer an issue now that the locking is done in the event loop.
While I don't have full confidence that it makes sense for the Drop implementation to behave
this way, this is nonetheless a significant improvement. The previous behavior led to
inconsistent state, panics, and event loop breakage, along with not actually destroying the
window.
This additionally makes the assumption that users don't need Focused or CursorLeft events
for the destroyed window, as Closed is adequate to indicate unfocus, and users may not
expect to receive events for closed/dropped windows. In my testing, those specific events
were sent immediately after the window was destroyed, though this sort of behavior could be
WM-specific. I've opted to explicitly suppress those events in the case of the window no
longer existing.
The `CursorMoved` events that are used to send position updates alongside `Focused` and
`CursorEntered` events were using incorrect values for the window ID. This is a direct
result of the X11 backend being hard to understand, as those values came from variables in
the top-level scope of the function, which one would assume to be valid throughout the
entirety of their scope. In reality, their validity is dependent on the event belonging to
the `XEvent` union, so very surprising things can happen if those variables are read in the
case of XInput2/XKB/etc. events. To prevent future accidents, the aforementioned variables
have been removed, and are now defined per-event instead.
Additionally, the `CursorMoved` event sent alongside `Focused` now uses the correct device
ID; it previously used the ID of a master keyboard, but now uses the ID of the pointer
paired to that keyboard. Note that for those using multi-pointer X, the correctness of this
ID is dependent on the correctness of the window manager's focus model.
Previously, the maximization hints were being sent as two separate client messages: one for
horizontal, and one for vertical. That resulted in the window only being maximized
horizontally (at least with my WM). The corrected client message sets both of these hints at
once.
In the process of implementing that, the relevant components were refactored to use the util
module, as we gradually move towards a hopeful future of a more readable X11 backend.
* Update mouse pos after cursor enter event
* Update mouse position on windows focus
* Send device_id
* Update other device id
* Fix windows import
* Remove deque for vec
* Just send event
* Use correct push_back method
* Push correct event
* Explicit mouse-related DeviceEvents
This makes the API more intuitive for common use-cases and allows us
to better propagate platform knowledge of motion semantics.
* Improve event naming consistency
* Clarify axis event forwards-compatibility
* Rename WindowEvent::MouseMoved/Entered/Left to CursorMoved/...
This emphasizes the difference between motion of the host GUI cursor,
as used for clicking on things, and raw mouse(-like) input data, as
used for first-person controls.
* Add support for windows and OSX, fix merging
* Fix warnings and errors on Linux
* Remove unnecessary breaking changes
* Add MouseWheel events to windows and OSX
* Fix bad push call.
* Fix docs, naming, and x11 events
* Remove mutability warning
* Add changelog entry
* Don't use UNIX_BACKEND in Window2::new
* Move get_available_monitors and get_primary_monitor to EventsLoop
* Remove UNIX_BACKEND
* Restore choosing the Linux backend
* Return a XNotSupported for new_x11()
* Fix fullscreen example
When X's evdev input module is configured to emulate scroll events (as
used with e.g. trackpoints), it generates non-emulated scroll button
presses and does not generate motion events. This is contrary to the
behavior of all other hardware I've tested, and contrary to the
behavior of libinput, but nonetheless should be supported.
