EglWaylandBackend gains support for creating textures from a
BufferInterface. At the same time it loses the possibility to use
the Xcb shm extension to load the texture. That is Xwayland is
required.
In order to support it in a better way the WindowPixmap is passed
to the Texture for loading and updating. Which is then passed to the
backend specific implementation.
Effects currently modify the matrices by reading back the uniform
values from the shader they assume will be used to paint the window,
set new values for the uniforms, call paintWindow(), and then restore
the uniforms to their previous values.
This is both slow and unreliable, and will not work with dynamically
generated shaders.
This patch solves the problem by putting the matrices in
WindowPaintData and making it the responsibility of the paintWindow()
implementation to set the uniforms.
The DecorationShadow supports the concept of sharing shadows between
multiple Decorations and this is supported by our Shadow class. But
still the created GLTextures were not shared.
This change introduces a sharing mechanismn for all GLTextures created
for a DecorationShadow. Thus if two decorations share the same
DecorationShadow they will also share the same GLTexture.
Register a callback function for debug output when KHR_debug or
ARB_debug_output is supported.
Only error messages and warnings about undefined behavior are
enabled in release builds, while all debug messages are enabled
in debug builds.
Move the buffer-swap-pending state from the compositing backends to
the Compositor class. The Compositor is the only class that needs to
access the state, and this way it to do it without calling through
a chain of virtual functions. This commit adds two new functions to
Compositor; aboutToSwapBuffers() and bufferSwapComplete(). The
backends call these functions to set and reset the buffer-swap-pending
state.
This commit also renames a number of functions and variables to make
their meaning clear.
The act of promoting the contents of the back buffer to become the
contents of the front buffer is referred to as posting the buffer,
presenting the buffer, or swapping the buffers; rendering the buffer
is what paintScreen() does.
Create and import X sync fences into GL and use them to synchronize
the kwin command stream with the X command stream.
This prevents damaged windows from being composited by kwin before
the rendering that triggered the damage events have finished on
the GPU.
Requires GL_EXT_x11_sync_object.
Tested-by: Marco Martin <notmart@gmail.com>
This adds a SceneOpenGL::Texture::load(..., xcb_visualid_t) overload,
and uses it to bind window pixmaps to textures.
By taking the RGBA masks in the visual into account when choosing an
FBConfig for the GLXPixmap, we are able to disambiguate formats that
have the same depth, such as GL_RGB10_A2 and GL_RGBA8.
The Renderer gets reparented to the Deleted. While passing it to
the Deleted the Scene's implementation can ensure that the buffers
are up to date. After passing to Deleted it's no longer allowed to
call the render method.
NOTE: this is not working completely yet, lots of code is still ifdefed
other parts are still broken.
The main difference for the new decoration API is that it is neither
QWidget nor QWindow based. It's just a QObject which processes input
events and has a paint method to render the decoration. This means all
the workarounds for the QWidget interception are removed. Also the paint
redirector is removed. Instead each compositor has now its own renderer
which can be optimized for the specific case. E.g. the OpenGL compositor
renders to a scratch image which gets copied into the combined texture,
the XRender compositor copies into the XPixmaps.
Input events are also changed. The events are composed into QMouseEvents
and passed through the decoration, which might accept them. If they are
not accpted we assume that it's a press on the decoration area allowing
us to resize/move the window. Input events are not completely working
yet, e.g. wheel events are not yet processed and double click on deco
is not yet working.
Overall KDecoration2 is way more stateful and KWin core needs more
adjustments for it. E.g. borders are allowed to be disabled at any time.
The left and right border images are rotated 90° before they are
uploaded into the atlas texture. The images are separated by a row
of transparent texels to minimize artifacts from oversampling.
With this change kwin renders the whole decoration with a single
call to glDrawArrays().
KWin already has a de facto OpenGL 2 dependency through QML. Combined
with the fact that the OpenGL 1 backend is basically unmaintained and
also unused, it's better to remove it for the new major release.
This change includes:
* Removal of cmake option KWIN_BUILD_OPENGL_1_COMPOSITING
* Removal of KWIN_HAVE_OPENGL_1 compile option and all code
ifdef'ed with it (partially removal of if-else constructs)
* Removal of CompositingType::OpenGL1Compositing (flags are kept
as a core flag should get introduced)
* Driver recommendation for OpenGL1Compositing changed to XRender
(should be evaluated whether the drivers can provide GL2)
* Removal of configuration option "GLLegacy"
* Removal of fooMatrix function in kwinglutils
* Removal of ARBBlurShader
* Removal of legacy code path in GLVertexBuffer
* Removal of GLShaderManager::disable
* if-blocks with ShaderManager::instance()->isValid() removed
REVIEW: 116042
Only the X based Scenes need an overlay window, so the Compositor doesn't
need to check for it in the Wayland case.
OverlayWindow is moved from OpenGLBackend to the sub classes which need
to provide it.
The egl wayland backend registers for the callback for a rendered frame.
This allows to throttle KWin's compositor so that we don't render frames
which wouldn't end up on the screen.
For this the Scene provides a method to query whether the last frame got
rendered. By default this returns true in all backends. The Egl Wayland
backend returns true or false depending on whether the callback for the
last frame was recieved.
In case the last frame has not been renderd when performCompositing is
tried to be called, the method returns just like in the case when the
overlay window is not visible. Once the frame callback has been recieved
performCompositing is invoked again.
The pure virtual methods windowAdded, windowClosed, windowDeleted and
windowGeometryShapeChanged had identical implementations in both XRender
and OpenGL scene. They were accessing the hash with Scene::Windows which
is nowhere else used except for creating the stacking order in ::paint.
The implementations are moved to the base class, the only Scene specific
code is a pure virtual factory method to create the Scene window. This
already existed in SceneOpenGL to create either a SceneOpenGL1 or 2
window.
Also the hash of windows is a Scene private member now and the creation
of the stacking order is provided by a method, so that the Scene sub
classes do no longer need to access the stacking order at all.
REVIEW: 111207
Instead of having the Shadow factory method check the compositor type and
do the decision which Shadow sub class to create, a pure virtual method in
Scene is called which returns the specific Shadow sub class instance.
Instead of having the EffectFrameImpl check the compositor type and do
the decision which Scene::EffectFrame to create, a pure virtual method
in Scene is called which returns the specific Scene::EffectFrame.
Allow prepareRenderingFrame() to return a region that will be
repainted in addition to the damaged region.
Pass both the damaged region and the repainted region, which
may be larger, as parameters to endRenderingFrame().
With QtQuick2 it's possible that the scene graph rendering context either
lives in an own thread or uses the main GUI thread. In the latter case
it's the same thread as our compositing OpenGL context lives in. This
means our basic assumption that between two rendering passes the context
stays current does not hold.
The code already ensured that before we start a rendering pass the
context is made current, but there are many more possible cases. If we
use OpenGL in areas not triggered by the rendering loop but in response
to other events the context needs to be made current. This includes the
loading and unloading of effects (some effects use OpenGL in the static
effect check, in the ctor and dtor), background loading of texture data,
lazy loading after first usage invoked by shortcut, etc. etc.
To properly handle these cases new methods are added to EffectsHandler
to make the compositing OpenGL context current. These calls delegate down
into the scene. On non-OpenGL scenes they are noop, but on OpenGL they go
into the backend and make the context current. In addition they ensure
that Qt doesn't think that it's QOpenGLContext is current by calling
doneCurrent() on the QOpenGLContext::currentContext(). This unfortunately
causes an additional call to makeCurrent with a null context, but there
is no other way to tell Qt - it doesn't notice when a different context
is made current with low level API calls. In the multi-threaded
architecture this doesn't matter as ::currentContext() returns null.
A short evaluation showed that a transition to QOpenGLContext doesn't
seem feasible. Qt only supports either GLX or EGL while KWin supports
both and when entering the transition phase for Wayland, it would become
extremely tricky if our native platform is X11, but we want a Wayland
EGL context. A future solution might be to have a "KWin-QPA plugin" which
uses either xcb or Wayland and hides everything from Qt.
The API documentation is extended to describe when the effects-framework
ensures that an OpenGL context is current. The effects are changed to
make the context current in cases where it's not guaranteed. This has
been done by looking for creation or deletion of GLTextures and Shaders.
If there are other OpenGL usages outside the rendering loop, ctor/dtor
this needs to be changed, too.
According to massif just creating the ColorCorrection costs 2 MiB, so
it makes sense to not create it if it is not used.
REVIEW: 111604
DIGEST: Smaller memory footprint in KWin 4.11.2 if color correction is disabled
This backend is able to composite on a Wayland surface instead of an X11
overlay window. It can be considered as a prototype for a Wayland session
compositor.
For texture from X11 pixmap the backend uses XShm. This is far from
optimal, but the KHR_image_pixmap extension is not available in Mesa's
Wayland backend. It's a temporary solution till we have XWayland and
texture from Wayland buffer.
To use this backend one needs to specify the environment variable
KWIN_OPENGL_INTERFACE with "egl_wayland". In future KWin should probably
use this backend if the Wayland display env variable is defined.
To use this setup:
1. Have a normal X-Server running on e.g. VT7
2. Start Weston on VT1
3. Start a terminal on Weston
4. start KWin with:
DISPLAY=:0 KWIN_OPENGL_INTERFACE=egl_wayland kwin --replace &
This should map a Wayland surface to Weston showing the content of the X
setup. At the moment it's not yet possible to interact with the surface
as input events are not yet recieved in the backend.
There are still a lot of limitations as documented in the code.
Cross fading with previous pixmap is achieved by referencing the old
window pixmap. WindowPaintData has a cross-fade-factor which interpolates
between 0.0 (completely old pixmap) to 1.0 (completely new pixmap).
If a cross fading factor is set and a previous pixmap is valid this one
is rendered on top of the current pixmap with opacity adjusted. This
results in a smoother fading.
To simplify the setup the AnimationEffect is extended and also takes care
about correctly (un)referencing the previous window pixmap. The maximize
effect is adjusted to make use of this new capabilities.
Unfortunately this setup has a huge problem with the case that the window
decoration gets smaller (e.g. from normal to maximized state). In this
situation it can happen that the old window is rendered with parts outside
the content resulting in video garbage being shown. To prevent this a set
of new WindowQuads is generated with normalized texture coordinates in
the safe area which contains real content.
For OpenGL2Window a PreviousContentLeaf is added which is only set up in
case the crass fading factor is set.
REVIEW: 110578
The behavior for creating a pixmap for a window is moved from Toplevel
into a dedicated class WindowPixmap. Scene::Window holds a reference to
this class and creates a new WindowPixmap whenever the pixmap needs to be
discarded. In addition it also keeps the old WindowPixmap around for the
case that creating the new pixmap fails. The compositor can in that case
use the previous pixmap which reduces possible flickering. Also this
referencing can be used to improve transition effects like the maximize
windows effect which would benefit from starting with the old pixmap.
For XRender and OpenGL a dedicated sub-class of the WindowPixmap is
created which provides the additional mapping to an XRender picture and
OpenGL texture respectively.
BUG: 319563
FIXED-IN: 4.11
REVIEW: 110577
Reimplement performPaint() in SceneOpenGL1Window and SceneOpenGL2Window.
The roles between begin/endRenderWindow() and performPaint() are now
reversed; performPaint() contains the specialized code while begin/
endRenderWindow() contains the shared code.
This reduces the state churn in the OpenGL2 backend from the repeated
calls to prepare/restoreStates().
