GitLab

GIT_SILENT

GIT_SILENT

GIT_SILENT

It's always emitted after the execution of the current function.

Also it breaks kwin's modifier_only_shortcut_test.

Fixes kwin's modifier_only_shortcut_test execution. Also it makes sense
that it's emitted eventually.

GetSessionByPid is broken especially when the calling process is in a
different cgroup to the session object.

Rather than using the magic "/auto" object path, calling GetSession with
the argument "auto" seems to work on even distros with old bases and is
the smallest code change.

BUG: 433364

This puts the greeter in the unmanaged layer within kwin, which is
what's expected in kwin tests.

On Wayland the virtual keyboard comes from the compositor.

Since the greeter has an exclusive zone of 0, it can be resized to
accommodate for other layer shell surfaces, for example panels.

Every view must have an exclusive zone of -1 so it fills the entire
screen and there are no gaps.

Using the PlasmaShell protocol has all sorts of problems because
placement is decided by the client at a position level rather than by
the application.
This adopts the layer shell and enforces the surface to be full screen
on the correct output and the compositor should be responsible for
placing it correctly.

If you want git blame to ignore formatting revisions run:
git config blame.ignoreRevsFile .git-blame-ignore-revs

It serves no purpose

As discussed in https://mail.kde.org/pipermail/plasma-devel/2021-February/119329.html

GIT_SILENT

Before KSldApp's constructor we call QObject::QObject. This calls
setParent which triggers a ParentChangedEvent on the virtual from the
constructor (boo!)

At this point our constructor hasn't run, so m_globalAccel isn't
initiliased.

Arguably this is a false positive as it always evaluates to the same
thing, but we don't want to be reading uninitialised memory.

This makes it easier to build against both Qt5 and Qt6

GIT_SILENT

GIT_SILENT

GIT_SILENT

GIT_SILENT

Screen geometry changed events could end up connected to the wrong
views, or not at all. This could cause greeters not to be visible (and
as such desktop contents visible), greeters at the wrong size or
position and/or overlapped.

We map Qt screens() by index to our own list of m_views, but Qt is free
to reorder screens, and when screens are removed, what used to be e.g.
index 0 could be removed, leading to connect() pointer to pointer
mappings being incorrect. Since we also set geometry on add/remove this
can lead to us changing to the correct geometry then back to geometry
that is not covered by active screen, essentially making the greeter
invisible, and no valid greeter visible.

Since there is no uniqueness constraint for lambda connections, we can
also end up with multiple screens updating different geometries on the
same view, or repeated signals from the same screen.

Fix this by only connecting screens once, and having the receiving
lambda resolve the index to index mapping so connections are persistant
with the life of the screen.

BUG: 347934

Make KScreensaverSettings Singleton as it inherits a singleton to avoid multiple parents singleton instanciation

Make quit a queued connection to give the QML part time to eventually execute code connected to Authenticator::succeeded if any. All such connections would need to be immediate

The replace method of the Controls 1 stack view is not docuemnted and apparently
does not return the item.
BUG:429290
FIXED-IN:5.20.4

It can happen that we connect to a logind where the session should
already be locked. So lock it accordingly.