This is a workaround fo QTBUG-34898 and affects the VirtualBox driver
and SandyBridge Mobile. OpenGL compositing is just not possible and
crashes as soon as there is anything rendered with QtQuick. This change
should be reverted once the Qt bug is fixed.
To nevertheless use OpenGL one can as always use the KWIN_COMPOSE env
variable, though this will result in crashes. An alternative is to set
QT_OPENGL_NO_SANITY_CHECK which forces Qt into using the threaded
rendering. At least for Sandybridge this seems to be a workable solution
as it's only causing flickering in fullscreen and KWin doesn't use any
fullscreen QtQuick elements.
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.
Instead of looping over every quad for every grid cell and checking
for possible intersections, loop over the quads once and compute
the top left corner of the first intersecting grid cell. Then loop over
all the intersecting cells from that point and create the sub quads.
This new algorithm also preserves the order of the quads in the
original list.
NOBODY should use org.kde.kded5 but instead directly the service
register by its own modulem, but for now let's just switch to .kded5
so we can get this running fast.
Only needs:
* Qt5::DBus for the communication with color correction
* Qt5::X11Extras for access to XCB, pulls in needed Qt5::Gui
* XCB component XCB for xcb calls in kwinglobals.h
It's basically a run of the port-cmake.sh script in here, mostly the changes
are the following:
- Using KF5::* targets
- Using the proper macros, following recent developments in frameworks
As KWin indirectly uses Qt's OpenGL through QtQuick we need to ensure
to not mix OpenGL and OpenGLES. So we have to built KWin only against
OpenGL if Qt is built against OpenGL and we have to built KWin only
against GLESv2 if Qt is built against GLESv2.
This means the kwin_gles binary is no more. There is only kwin which
either links GL or GLESv2.
also work around broken fbo texture clearing on fglrx
so far supports FBO/glClear and resorts to glTexSubImage2D
if the fbo cannot be created or is (in case of fglrx)
known to break, resort to glTexImage2D loading of an
argb array of zeros
BUG: 323065
FIXED-IN: 4.11.2
REVIEW: 112526
Port from kFatal to qFatal - kFatal is still in kde4support,
but lacking debug area support arguably isn't that big a
deal since fatals can't be turned off anyway.
CCMAIL:mgraesslin@kde.org
On GLES, check for OES_texture_3D extension for color correction
Remove a block of ugly hack code that was supposedly needed to build
with OpenGL ES.
Convert the lookup texture data to uint8 on OpenGL ES before sending it
via glTexImage3D, because uint16 is not supported.
Check if the shaders have been reinitialized successfuly when trying to
activate color correction, prevent black screens when there are issues
with the shaders.
BUG: 315419
REVIEW: 111225
(cherry picked from commit 68c68ee3c2b54f968c4d8275f1e8a2e0ccc90dd7)
* "" needs to be wrapped in QStringLiteral
* QString::fromUtf8 needed for const char* and QByteArray
* QByteArray::constData() needed to get to the const char*
Qt 5 only supports raster which means our pixmaps are always non native,
so we don't need the Extension information any more and can drop all
special code handling for mapping a native QPixmap to an X11 pixmap.
This method returns true when glMapBufferRange() is likely to perform
worse than glBufferSubData() when updating an unused range in a buffer
object.
This is the case with the NVIDIA driver, where glMapBufferRange()
will force thread serialization. The driver tracks which ranges of
the buffer are in use, so calls to glBufferSubData() should not
cause a pipeline stall.
Assume that the default framebuffer has the same dimensions as the screen.
By not quering the dimensions of the viewport we don't risk serialization
in drivers that use threaded dispatch.
This reverts commit 23dff966437bb664a2ffdb3f7957ef39978f5fad.
Using QVector is not a win when effects such as wobbly windows are
active, due to the realloc overhead. So revert this change for now.
Eg. gtk+ alters the modality after mapping and
before unmapping the window.
Therfore the former implementation ahd a wrong idea
about the modality until the window was activated and
again had a wrong idea when the dialog closed, keeping
the main client dimmed.
Modality changes at runtime are uncommon but legal and can
happen anytime.
BUG: 321340
FIXED-IN: 4.11
REVIEW: 111154
Prevents the possiblity of using shaders modified for color correction
without valid data from KolorManager. If that happened, everthing
blacked out.
Now the color correction shaders are enabled only after successfuly
contacting KolorManager.
The issue was highlighted after ab7e228d.
BUG: 321217
it seems an animation can be triggered and the resp.
window deleted in the first event cycle (before the deleted
signal is bound) - so we add an initialization flag to ensure
the binding happens before the first animation is added
BUG: 320562
FIXED-IN: 4.11
REVIEW: 110872
Just like Opacity for transforming only the decoration opacity.
It's an ABI break as the new enum value had to be included as anonther
non-float base value.
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
This avoids the overhead of allocating each WindowQuad on the heap
when appending items to the list, and also ensures that the quads
are continuous in memory.
This reduces the size of the geometry that needs to be uploaded by
one-third, and allows kwin to take advantage of the post-transform
cache in the GPU.
Expose bindArrays(), unbindArrays() and add a draw() method that takes
an offset and a count. This makes it possible to upload geometry, call
bindArrays(), and then call draw() multiple times to draw different
subsets of the uploaded geometry.
Split WindowQuadDecoration into WindowQuadDecorationLeftRight
and WindowQuadDecorationTopBottom.
This simplifies the code in SceneOpenGL::Window::paintDecoration().
Do not schedule repaints in AnimationEffect if there are no animations
going on. If an animation is waiting for starting or kept after ending
the visual appearance is no longer changed, so no repaint is needed.
REVIEW: 110795
Unlike makeArrays() this function writes into a pre-allocated array,
and takes a matrix that's used to transform the texture coordinates.
This allows this function to handle coordinates for rectangular
textures correctly.
Note that unlike the previous commit, this doesn't fix texture coordinates
for rectangular textures. That case cannot be handled correctly without
knowing the dimensions of the texture.
This method returns a matrix that transforms normalized or un-normalized
texture coordinates, taking the texture target and y-inversion flag into
account.
KWin always updates the array buffer binding before it calls GL functions
that reference it, so there is never any need to reset it.
This should eliminate half the calls to glBindBuffer() while painting
the scene.
These methods make it possible to write directly into the buffer object
when building vertex arrays.
If the buffer object cannot be mapped, the map() method will return
a pointer to local memory which will be submitted to the buffer object
with glBufferData() when unmap() is called.
This overload makes it possible to upload data of an arbitrary size and
type into the buffer object. The intent is for this method to be used
to upload interleaved vertex data.
This commit also adds setVertexCount() and setAttribLayout().
The rationale for decoupling attribute specification from data uploading
is that the attribute formats and layout change less frequently than
the vertex data.
The vertex count is also specified using a separate function to enable
the caller to upload data for multiple draw calls at the same time.
Store the formats as an array in GLVertexBufferPrivate.
This simplifies the code for enabling the generic vertex arrays,
and also makes it easier to add new arrays.
Consolidate the code for binding and unbinding the vertex arrays into
two new methods called bindArrays() and unbindArrays() respectively.
This patch also removes the three paint implementations, since the only
difference between them is the code that sets up the arrays. The actual
painting code is moved into GLVertexBuffer::render(), which uses the
new methods to bind and unbind the arrays.
In case OpenGL ES 3 is provided by the driver we can use the GLSL 1.40
shaders as GLSL 300 ES shaders. The #version declarative is rewritten in
such a case.
REVIEW: 110590
Only the subset of functions available in core contexts is resolved,
except for glGetnTexImageARB() and glGetnUniformivARB(), which are
not used by kwin.
Instead of setting the function pointers to NULL when the extension isn't
supported, kwin provides its own implementations that call the non-robust
versions of the functions. This is so callers don't have to check if the
extension is supported before calling the functions.
Allocate enough space to hold the geometry for multiple draw calls,
and use glMapBufferRange() to gradually fill the buffer. Once the
data store is full, it's orphaned and a new one is allocated.
Store the vertex positions and texture coordinates in the same buffer
object. This saves one buffer allocation in every setData() call.
The attributes are also interleaved as they are uploaded into the buffer
to maximize locality of reference.
This patch adds a link() function, along with bindAttributeLocation()
and bindFragDataLocation().
These functions must be called after creating the program, but before
linking it.
A new ExplicitLinking flag must be passed to the constructor to prevent
automatic linking. This is to keep existing code working without
modifications.
