The resize handler with a window resize request causes oscillations between two window sizes (presumably the inner and the outer). While it's not clear what part of the stack causes it, checking `window.get_inner_size()` already presents the correct size.
Either way, changing the same property in its change handler is a recipe for feedback loops as this one and should be avoided whenever possible, and then carefully fortified too.
This commit removes some clippy warnings (to advance #45) by doing the following:
- replace usage of `mem::uninitialized()` with `MaybeUninit`
- replace usage of `nix::libc::{uint64_t, int32_t}` with `{u64, i32}`
- replace functions inside of `Option::ok_or` with `Option::ok_or_else`
- replace functions inside of `Result::unwrap_or` with `Result::unwrap_or_else`
- replace occurrences of pass-by-reference with pass-by-value when
appropriate
- replace unused variables in pattern-matching with wildcards
- replace `match` expressions that have only one case with `if let`
expressions
- replace UpperCamelCase names of consts with SCREAMING_SNAKE_CASE
- remove `clone()` on types that implement Copy
- remove redundant imports
- remove `fn main()` from doctests
- remove let binding for variables that are returned afterwards
* Rename priviledged to privileged in backend/drm/mod.rs
* Fix a number of typos in comments
* Fix typos in string literals
* Wrap identifiers with backticks in doc comments
* Spelling and capitalisation in doc comments
* Use XWayland in comments
This seems to be the standard capitalisation, even though the executable
is called `Xwayland`.
* Use Glium instead of glium in comments
* Use DRM and API in comments
* Fix remaining occurrence of 'priviledged'
* Reformat code to appease Travis' rustfmt
This event poses issues regarding its semantic (what space are the
coordinates in?) and arguably won't be used anyway. So let's remove it
for now, it can still be added afterwards if it appears to be really
needed.
Unify winit with the other backends in only ever dealing with physical
sizes. All dimensions and coordinates are provided in the physical space
(transformed by the dpi factor provided by winit), and the user is
responsible for dealing with the dpi scaling if they want.
A compositor using it can deal with it either by following the scaling
transmitted by the backend or using some value retrieved by other means
(like a command line argument, like weston does).
This means that a non-hidpi aware implementation will appear very small
on an hidpi screen, but not broken. This is now the current state of
anvil.
This also changes the WinitEventsHandler trait to merge the resized()
and hidpifactorchanged() callbacks into a single one, providing the
physical size and the dpi factor whenever either changes.
- Decouple DrmBackend and DrmDevice through Weak references
- Move EGL-WlBuffer related functionality into it's own trait
- Impl new trait by every struct that can theoretically bind the display although it is no rendering target (no EGLGraphicsBackend), e.g. DrmDevice
- Move texture binding into own struct `EGLDisplay` created by `bind_wl_display` that can be passed around freely
- Add device num to DrmDevice for HashMap storage
- Fixup and enable acceleration on drm and udev examples made possible by the previous changes.
Victor Berger
Victor Brekenfeld
Chandler Newman
Sergey Smirnykh
Dorota Czaplejewicz
nycex
Colin Benner
Jonas Platte
Lucas Timmins
Colin Benner