Summary:
Qt's metaobject is rather sensitive with scope resolution.
Foo::Bar and Bar don't always match to a Qt metaobject, even if they
refer to the same thing to a compiler. Here we register
X11Compositor::SuspendReason but Q_ARG uses SuspendReason and they don't
match. This leads to a runtime failure where the method isn't invoked.
Rather than fixing metaobject usage, port the whole thing to lambdas
which does better compile time checking and is generally nicer to read.
BUG: 412353
Test Plan:
Ran xprop to block compositing. Compositing was blocked.
Grepped source code for Q_ARG use
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D24244
Summary:
Because KWin is a very old project, we use three kinds of null pointer
literals: 0, NULL, and nullptr. Since C++11, it's recommended to use
nullptr keyword.
This change converts all usages of 0 and NULL literal to nullptr. Even
though it breaks git history, we need to do it in order to have consistent
code as well to ease code reviews (it's very tempting for some people to
add unrelated changes to their patches, e.g. converting NULL to nullptr).
Test Plan: Compiles.
Reviewers: #kwin, davidedmundson, romangg
Reviewed By: #kwin, davidedmundson, romangg
Subscribers: romangg, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23618
Summary:
Compositor::hasScene() is redundant. Depending on use case, it can be
replaced by checking either m_state or Compositor::scene().
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23744
Summary: Less likely to happen but still we need to handle this case.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23611
Summary: Overlay windows is an X11 thing.
Test Plan: Compiles.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23608
Summary:
Switch to Q_ASSERT in order to make code a bit more consistent. We have
places where both assert and Q_ASSERT are used next to each other. Also,
distributions like Ubuntu don't strip away assert(), let's hope that
things are a bit different with Q_ASSERT.
Test Plan: Compiles.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: romangg, davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23605
Summary:
Any call made to a virtual method in constructor/destructor of a base
class won't go to a derived class because the base class may access
uninitialized or destroyed resources.
For example, let's consider the following two classes
class Base {
public:
Base() { foo()->bar(); }
virtual ~Base() { foo()->bar(); }
virtual Foo* foo() const { return nullptr; }
};
class Derived : public Base {
public:
Derived() : mFoo(new Foo) {}
~Derived() override { delete mFoo; }
Foo* foo() const override { return mFoo; }
private:
Foo* mFoo;
};
When an instance of Derived class is created, constructors will run in
the following order:
Base()
Derived()
It's not safe to dispatch foo() method call to Derived class because
constructor of Derived hasn't initialized yet mFoo.
Same story with destructors, they'll run in the following order:
~Derived()
~Base()
It's not safe to dispatch foo() method call in the destructor of Base
class to Derived class because mFoo was deleted.
So, what does that weird C++ behavior has something to do with KWin? Well,
recently Compositor class was split into two classes - WaylandCompositor,
and X11Compositor. Some functionality from X11 doesn't make sense on
Wayland. Therefore methods that implement that stuff were "purified," i.e.
they became pure virtual methods. Unfortunately, when Compositor tears
down it may call pure virtual methods on itself. Given that those calls
cannot be dispatched to X11Compositor or WaylandCompositor, the only
choice that C++ runtime has is to throw an exception.
The fix for this very delicate problem is very simple - do not call virtual
methods from constructors and the destructor. Avoid doing that if you can!
This change moves Compositor::updateClientCompositeBlocking to X11Compositor
so it longer has to be a virtual method. Also, it kind of doesn't make sense
to keep it in base Compositor class because compositing can be blocked only
on X11.
BUG: 411049
Test Plan: KWin no longer crashes when running kwin_x11 --replace command.
Reviewers: #kwin, romangg
Reviewed By: #kwin, romangg
Subscribers: anthonyfieroni, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23098
Summary:
With the split of the Compositor class in X11 and Wayland it is unnecessary
to check in X11Compositor if compositing is required since on X it never is.
Test Plan: Compiles.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D23032
Summary: Replaces foreach loops with modern for loops and improve code style overall.
Test Plan: Auto tests pass as before and manually in X and Wayland sessions.
Reviewers: #kwin
Subscribers: zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D23011
Summary:
This patch is a first take at splitting up of the Compositor class into
Wayland and X11 child classes.
In this first patch we mostly deal with setup and teardown procedures.
A future goal is to further differentiate the compositing part itself too.
Test Plan: Manually X from VT and Wayland nested. Autotests pass.
Reviewers: #kwin
Subscribers: sbergeron, anthonyfieroni, zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22195
Summary:
Replace the several internal state booleans of Compositor with a single
enum to register the current state of the compositor. We register four
states of starting, started, stopping and stopped.
The goal is to replace the several different conditionals when starting
and stopping the compositor with a single well defined flow.
There are currently still some ugly conditionals and some replaced ones
might need some more work.
Test Plan: Manually in X and Wayland. Relevant auto tests pass.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22277
Summary:
To streamline Compositor code more remove the composite reset timer. The two
times it was used we can either use a singleshot timer instead or connect the
call to a different signal in the X11 backend.
Long term goal is to have a well structured init of the Compositor such that
we can call directly instead.
Test Plan: Manually in X and Wayland nested session.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22270
Summary:
This removes the restart function of the Compositor class and renames the
internal reinitialize function.
Instead of the restart function reinitialize can be called. Reading again
the settings in this case is fine, since it is done rarely. This reduces
the code complexity.
Test Plan: Manually on Wayland and X. 100% autotests pass.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: davidedmundson, zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22225
Summary:
The check is used to limit settings reload. But we can afford to reload
settings in case the compositing gets toggled since this happens rarely.
Removing the check reduces code complexity.
Test Plan: Manually in X and Wayland.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22220
Summary:
As a preparation for further changes clean up the Compositor
class. First step is to use the new slot syntax and rename
some of the slots.
Includes some other minor code style improvements to the class
as well.
Test Plan: Manually in X and Wayland nested session.
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: davidedmundson, zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D22218
Summary:
Declare it in the source file for internal usage of ApplicationX11 and
Compositor classes. Additionally use modern style connect and do other
code style updates.
Test Plan: Manually in X session and Wayland windowed.
Reviewers: #kwin, zzag
Subscribers: davidedmundson, sbergeron, zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D21655
Summary:
Compositing today is ubiquitous. There is no reason to keep compositing
specific functions of Toplevel, Client and Workspace classes in the
composite.cpp source file. Instead let these definitions be separated.
Test Plan: Compiles
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Maniphest Tasks: T11071
Differential Revision: https://phabricator.kde.org/D21654
Summary:
KWin is not only an X11 compositing window manager, but also a Wayland
compositor. The Compositor class is used in both cases so in general it
would be nice to keep it as much as possible generic.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D19879
Summary:
Deleted has lifetime bounded to effects and scene.
We can't discard Deleted before EffectsHandler is destroyed because
effects that referenced it may call unrefWindow.
On the other hand, the fact that Deleted may outlive scene doesn't
make sense because Deleted exist purely for animation purposes and
nothing more.
This change arranges lifetime of Deleted so it's more reasonable.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18914
Summary:
Currently, each frame callback sent by KWin has the current time in
nanoseconds, but the protocol spec states that we have to send the time
in milliseconds. This is the reason why animations that are driven by
frame callbacks are too fast.
In addition to that, m_timeSinceStart isn't actually "time since start,"
it's rather accumulated duration of all painting cycles. If there is
something to draw and it takes quite a while to compose the scene, maybe
m_timeSinceStart will be close enough to the current time. So, it has
been replaced with QElapsedTimer, this makes the current time correct
and also simplifies code a little bit.
Test Plan: The triangle in weston-subsurfaces no longer spins very fast.
Reviewers: #kwin, romangg
Reviewed By: #kwin, romangg
Subscribers: romangg, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18656
Summary:
On Wayland we cannot switch from OpenGL to QPainter compositor as this
would break any running OpenGL application. KWin registers it's
EGLDisplay to Wayland and without OpenGL this doesn't make sense any
more. We are not able to render OpenGL buffers in the QPainter
compositor.
While it's theoretically possible to switch from QPainter to OpenGL it
doesn't make any sense for the same reason. Any running OpenGL
application would be using llvmpipe and could not be switched to proper
OpenGL.
This change stores the selected compositing type in Platform and the
implementations can use it to restrict the supported compositors. On X11
we don't need this, all other Platforms implement the restriction. Thus
it's no longer possible to switch the backends at runtime.
Test Plan:
Adjusted tests run, no runtime test as gui doesn't support
switching to QPainter anyway.
Reviewers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D19084
Summary:
Adds an autotest to show that KWin fails an assertion when a client tries to
resize a sub-surface.
Since it is the first autotest dealing with sub-surfaces explicitly additional
autotest helpers are introduced to allow that.
We also add a new signal in Compositor to spy on to know when the buffer swap
has been completed.
Test Plan:
Test fails as expected:
```
QFATAL : KWin::BufferSizeChangeTest::testShmBufferSizeChangeOnSubSurface() ASSERT: "image.size() == m_size" in file /home/roman/dev/kde/src/kde/workspace/kwin/platformsupport/scenes/opengl/abstract_egl_backend.cpp, line 394
FAIL! : KWin::BufferSizeChangeTest::testShmBufferSizeChangeOnSubSurface() Received a fatal error.
Loc: [Unknown file(0)]
Totals: 4 passed, 1 failed, 0 skipped, 0 blacklisted, 367ms
********* Finished testing of KWin::BufferSizeChangeTest *********
```
Reviewers: #kwin, zzag
Subscribers: zzag, graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18452
Summary:
Currently, KWin/Wayland crashes when the compositor is reinitialized.
The reason for that is ShellClient's DecorationRenderer gets destroyed
when the scene is already gone, thus there is no current OpenGL context.
Client works around that issue by destroying scene-specific DecorationRender
in finishCompositing. Such a workaround could be applied to ShellClient
as well, but it would make code more confusing because DecoratedClientImpl
also tries to destroy DecorationRenderer.
A better approach would be to notify DecoratedClientImpl that
compositing is about to be finished, so it can destroy the decoration
renderer when the scene is still alive. This not only fixes the
previously mentioned issue in ShellClient, but also makes code a little
bit tidier.
Test Plan:
Start Plasma on Wayland session, change any compositor settings (e.g.
animation speed).
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18921
Summary:
Code that destroys all scene windows when finishing compositing is not
readable. E.g. can you tell what this piece of code is doing without
looking into the source code of Scene::windowClosed?
foreach (Client * c, Workspace::self()->clientList())
m_scene->windowClosed(c, NULL);
This change intoduces removeToplevel(as well its counterpart) method to
the Scene class. The name of the new method much better describes what
we're doing.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18210
Currently, if a Wayland surface registers a frame callback but does
not send a damage event, the callback will not be serviced even after
KWin has completed a compositing cycle and is therefore ready to
render a new frame for the surface. Since some clients, including any
using NVIDIA's Wayland EGL implementation, wait for frame events
before preparing each frame this behavior can result in hangs if a
compositing cycle occurs between registering the frame callback and
sending the damage event.
Instead, in accordance with the Wayland specification, frame callbacks
should be serviced (via KWayland::Server::SurfaceInterface::frameRendered)
for all visible windows - not just those damaged since the last
compositing cycle.
Summary:
The SceneOpenGLES test was starting to randomly crash on my system. On
investigation I observed that there was a graphics reset ("Attempting to
reset compositing.") which triggered a delayed restart of the
compositor. This even was only processed while waiting for XWayland to
terminate. Which resulted in a crash in KWin::getXServerVersion as the
xcb connection broke.
It makes no sense to setup the compositor again during application
shutdown. Thus the dtors set a flag that they are terminating the
application and Compositor::setup is checking for it.
Test Plan: Test no longer crashes, although it goes through the crashing path
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D18015
Summary:
Compositing is suspended/finished in a very hard way fashion, effect
windows are destroyed without notifying effects about it.
AnimationEffect tries gracefully release deleted windows, but because
in some cases(like when suspending compositing) a deleted window can
be already destroyed, a segmentation fault can happen.
This change adjusts the order in which effect windows and effects are
destroyed, so AnimationEffect (and other effects) cannot access dangling
pointers.
BUG: 400788
FIXED-IN: 5.15.0
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D17311
Summary: We can get rid of the superfluous if/else ladder.
Reviewers: kwin
Subscribers: #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D10393
Summary:
If KWin/Wayland starts up with the QPainter compositor we force software
rendering in QtQuick. It doesn't make sense to try to use the OpenGL
renderer as that normally just results in an abort when trying to create
the QOpenGLContext.
Test Plan:
Run KWin with KWIN_COMPOSE=Q, tried to quick tile a window and
used Alt+Tab. Worked fine, didn't crash anymore.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8663
Summary:
With the new try of all compositor types supported there is an automatic
fallback from OpenGL to XRender/QPainter in case OpenGL setup failed.
So there is no need to invoke a method to do just that.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D8364
Summary:
The Compositor now tries to create a Scene not just once but every
type supported by the Platform till it finds one which works. The user's
configuration is only used as a preferred hint and tried first if the
platform supports it.
This brings as an advantage that on platforms such as framebuffer the
user does not need to specify which compositor to use: KWin uses QPainter
automatically.
Also we don't need to do the "translation" from XRender to QPainter any
more. XRender is not supported by any platform using QPainter, so the
user configuration is ignored anyway.
BUG: 350159
FIXED-IN: 5.12.0
Test Plan: Run on framebuffer, verified debug output.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D8363
Summary:
The Platform API is extended by a call to create the EffectsHandler. In
X11 standalone Platform a new EffectsHandlerImplX11 is added which
contains the X11 only parts of the EffectsHandler, such as grabbing the
X keyboard and the X11 mouse interception window.
The EffectsHandlerImpl gains some virtual methods for the parts which
are now done in the X11 specific implementation. In return we get rid of
lots of if-else structures checking for the operation mode.
Test Plan: Only compile tested.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7955
Summary:
Unfortunately a rather large change which required more refactoring than
initially expected. The main problem was that some parts needed to go
into platformsupport so that the platform plugins can link them. Due to
the rather monolithic nature of scene_opengl.h a few changes were
required:
* SceneOpenGL::Texture -> SceneOpenGLTexture
* SceneOpenGL::TexturePrivate -> SceneOpenGLTexturePrivate
* texture based code into dedicated files
* SwapProfiler code into dedicated files
* SwapProfiler only used in x11 variants
* Safety checks for OpenGL scene moved into the new plugin
* signal declared in SceneOpenGL moved to Scene, so that we don't need
to include SceneOpenGL in composite
Test Plan: Nested OpenGL compositor works
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7740
Summary:
Several of the subclasses are already derived from QObject.
The main reason is that the class should be moved out of KWin core in
order to move the OpenGL scene into a plugin. As Compositor calls into
the AbstractEglBackend to unbind the wayland display this creates a
problem which is easily solved by turning the AbstractEglBackend into a
QObject and connect to the signal emitted by the Compositor.
Test Plan: Compiles
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7669
Summary:
The compositor needs to claim the X11 compositor selection and redirect
the X11 windows. This of course only makes sense when having X11 support.
This change refactors the code so that if X11 support is missing the code
is not executed, but as soon as X11 support comes available the selection
gets claimed.
Also if the connection goes away the selection is deleted, though it
might be that this does not work as KSelectionOwner might call into xcb
and cause a crash. This needs to be tested once we start supporting
XWayland going away.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7504
This change is similar to D7232 and moves the scene_qpainter into a
dedicated plugin. Compared to the XRender case it's more complicated as
the platform plugins need to implement a platform specific backend.
The base implementation for this part used to be in scene_qpainter. As
the idea is to completly move it away from KWin core it would be point
less to still have the backend definition in KWin core, but it cannot
be in the scene plugin as otherwise all platforms need to link the
plugin.
To solve this a new platformsupport subdirectory is added which contains
the scene platform backend as a static library. For the OpenGL scene such
a static library will also be required.
Test Plan: SceneQPainter test still passes, nested compositor still works
Reviewers: #kwin, #plasma
Subscribers: plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D7259
Summary:
First step for loading the compositor Scenes through plugins. The general
idea is that we currently needlessly pull in all the Scenes although only
one will be used.
E.g. on X11 we pull in QPainter, although they are not compatible. On
Wayland we pull in XRender although they are not compatible.
Furthermore our current Scene creation strategy is not really fault
tolerant and can create situations where we don't get a compositor. E.g
on fbdev backend the default settings won't work as it does not support
OpenGL.
Long term I want to tackle those conceptional problems together:
we try to load all plugins supported by the current platform till we have
a scene which works. Thus on Wayland we don't end up in a situation where
we don't have a working compositor because the configuration is bad.
To make this possible the switch statement in the Scene needs to go and
needs to be replaced by a for loop iterating over all the available
scenes on the platform. If we go there it makes sense to replace it
directly with a plugin based approach.
So this is a change which tackles the problem by first introducing the
plugin loading. The xrender based scene (as it's the most simple one)
is moved into a plugin. It is first tried to find a scene plugin and only
if there is none the existing code is used.
Test Plan: Tested all scenes
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D7232
Summary:
The OverlayWindowX11 also inherits from X11EventFilter and performs
the filtering itself.
Test Plan: Compiles, not yet tested as I'm on Wayland
Reviewers: #kwin, #plasma
Subscribers: plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D7197
Introduce a method Workspace::markXStackingOrderAsDirty
Summary:
This method replaces the calls x_stacking_dirty = true in the code base
allowing for further refactoring of that functionality.
Remove roundtrip to XServer from Workspace::xStackingOrder
The method xStackingOrder is only used during a Compositor paint pass.
If the stacking order had changed, the method updated the stacking order
from X by performing a sync XQueryTree. With other words we had a round
trip to the X server directly in the paint pass.
This change rearchitectures this area by making better use of xcb. When
we notice that the stacking order changed and an XQueryTree is needed,
we directly send out the request. When xStackingOrder is finally called,
which normally happens a few milliseconds later, the reply is retreived.
In the worst case it still blocks, but in most cases the roundtrip is
gone.
If the stacking order changed again before accessing xStackingOrder the
running request is cancelled and a new request is issued. So whenever we
get into xStackingOrder it will have the current state.
The updating of the xStackingOrder is moved into a dedicated method and
xStackingOrder invokes it through a const_cast instead of operating on
mutable variables.
Test Plan: Normal system usage, no issues
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D6323
Summary:
While investigating a repaint issue I had a hard time to work with the
code and decided to change it to a more modern variant.
Instead of duplicating the same logic for each of the lists it's now
moved into a dedicated templated method which used std::any_of. Which is
exactly the algorithm we need: if there is any window with a non empty
repaints region it should return true.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3397
Summary:
With nouveau driver it can happen that KWin gets frozen when first trying
to render with OpenGL. This results in a freeze of the complete desktop
as the compositor is non functional.
Our OpenGL breakage detection is only able to detect crashes, but not
freezes. This change improves it by also added a freeze protection.
In the PreInit stage a thread is started with a QTimer of 15 sec. If the
timer fires, qFatal is triggered to terminate KWin. This can only happen
if the creation of the OpenGL compositor takes longer than said 15 sec.
In the PostInit stage the timer gets deleted and the thread stopeed
again.
Thus if a freeze is detected the OpenGL unsafe protection is written into
the config. KWin aborts and gets restarted by DrKonqui. The new KWin
instance will no longer try to activate the freezing OpenGL as the
protection is set.
If KWin doesn't freeze the protection is removed from the config as
we are used to.
Check for freezes for the first n frames, not just the first
This patch changes the freeze detection code to detect freezes in the
first 30 frames (by default, users can change that with the
KWIN_MAX_FRAMES_TESTED environment variable). This detects
successfully the freezes associated to nouveau drivers
in https://bugzilla.suse.com/show_bug.cgi?id=1005323
Reviewers: davidedmundson, #plasma, #kwin, graesslin
Reviewed By: #plasma, #kwin, graesslin
Subscribers: luebking, graesslin, kwin, plasma-devel, davidedmundson
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D3132
This fixes a regression found in TestSceneQPainter for restarting the
Compositor. The internal clients were never added to the Scene.
We cannot just call setupCompositing because there is a phase during
startup where setupCompositing already passes but the Compositor is not
yet fully initialized. Thus it could happen that it's called twice which
makes the Scene assert.
To solve that setupCompositing and finishCompositing are now overriden
in ShellClient and track whether they successfully called
setupCompositing.
