Nested classes inherit the visibility, so need to explicitly opt out again

This method is used to sync two outputs, previously, scale wasn't synced, now it is.

Summary:
When a monitor is scaled the "size" is not the same as the mode size, we
want things in global co-ordinates.

This is only relevant on Wayland, but on everything else scale is 1, so
this makes no difference.

Test Plan:
Plugged in normal monitor to laptop, generator had the correct co-ordiantes
of second monitor

Reviewers: #plasma, graesslin

Reviewed By: #plasma, graesslin

Subscribers: plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D5697

This should fix running Plasma in a windowed virtual machine, when the window is
resized, the mode list changes, and libksreen can't currently handle that.

Summary:
* make Output::modes() non CONSTANT, add modesChanged() signal
* compare the mode lists and set the new one
* queue an outputChanged signal when applied
* autotest for modelist changes
* update the mode list on RRNotify events

BUG:356864

Test Plan:
* for library part, autotests are added
* for xrandr backends, we can't sensibly autotest this :(

Reviewers: #plasma, mart

Reviewed By: mart

Subscribers: graesslin, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D3117

Summary:
This adds a scale property to KScreen::Output. By default, it's 1.0,
which means no scaling. The value is a qreal, and the conversion to int
for the backends that only understand int is made in the underlying
backend code. RFC. :)

Test Plan: autotests included

Reviewers: graesslin, davidedmundson

Reviewed By: davidedmundson

Subscribers: plasma-devel, davidedmundson

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D3312

This should fix running Plasma in a windowed virtual machine, when the window is
resized, the mode list changes, and libksreen can't currently handle that.

Summary:
* make Output::modes() non CONSTANT, add modesChanged() signal
* compare the mode lists and set the new one
* queue an outputChanged signal when applied
* autotest for modelist changes
* update the mode list on RRNotify events

BUG:356864

Test Plan:
* for library part, autotests are added
* for xrandr backends, we can't sensibly autotest this :(

Reviewers: #plasma, mart

Reviewed By: mart

Subscribers: graesslin, plasma-devel

Tags: #plasma

Differential Revision: https://phabricator.kde.org/D3117

This change introduces caching of backends in the backendmanager also
for in-process backends, along with proper memory management of the
backends.

I haven't entirely sorted where the Edid is set, this makes testing a
bit easier for now

This change introduces caching of backends in the backendmanager also
for in-process backends, along with proper memory management of the
backends.

I haven't entirely sorted where the Edid is set, this makes testing a
bit easier for now

Output::geometry() always reflects whatever is currently stored in
the Output (which can be different from what is the currently active
configuration). This was broken in the previous commit, because the
size() property is only updated when we get update from KScreen backend.

The API documentation for those properties has been updated to
explictly and clearly describe the difference in behaviour of those
properties.

CCMAIL: afiestas@kde.org

We were making some assumptions in Output::geometry like for example
assuimg that the resolution of the currentMode was actually the output
size which is not always the case, like for example in a case we were
already covering of vertical transformed output.

In order to fix this proplerly we are adding a size property to Output
that backends and fill with whatever value they want and we use it to
create the QRrect with return in Output::geometry.

QScreen: Whatever QScreen::screenGeometry().size() returns
Fake: Added a test for singleOutput, the rest is all the same
XRandR11: Whatever size has currentMode

XRandR: Whatever size CRTC reports.
Note that crtc is the one in charge of applying transformations so it
always reports correct geometry, even on scaled or vertical/landscape
outputs.

It's beyond my comprehension how I actually managed to mess
up the logic like that.

This seems to be the key point for many of the unexplainable crashes
that we've been seeing all over the place, especially in the KDED module.
When the ConfigMonitor is updating the properties one by one, the
change signals are being emitted immediatelly. So if someone is listening
to these signals, and will try to access some property that we have not
udpated yet, they will get inconsistent data (like empty mode list for
instance), leading to crashes.

With this patch we block signals emission on the output, collect all changes
that happened, and then emit the signals all at once after we updated all
the properties. This way applications can safely access any properties.

It's possible to disable fetching EDID by passing NoEDID option to GetConfigOperation
constructor.

The EDID retrieval method is a super hack for now, as it synchronously asks
BackendManager for backend, and then synchronously does a DBus call to the
backend to retrieve EDID. I don't know yet how to solve this properly: we
could either have flag in GetConfigOperation that would allow excluding EDID
from regular retrieval, or we need GetEdidOperation. *shrug*

The old Config::current() and Config::setConfig() methods were removed,
instead GetConfigOperation and SetConfigOperation classes are introduced
here. The emit finished() once the operation has completed. The operations
will auto-start and auto-delete.

The reason for having the async API is that BackendManager might need to start
the BackendLauncher first, and then still a DBus call is involved to get data
from the backend (which might also block for a moment). This would never be
acceptable to be called in blocking manner from main thread, hence nice simple
async API has been introduced.