When trying to connect to a network from gdm, it doesn't make sense to query
secrets from the gdm user since it's a system user.
Furthermore, gdm runs an isolated dbus-session per gnome-shell instance
(for multi-seat setups). Instead, gnome-keyring-daemon is started by systemd
and so it registers on the _main_ dbus session of the gdm user session.
Then, gnome-shell tries to dbus-activate another gnome-keyring-daemon on its
isolated bus, but gnome-keyring-daemon refuses to start as it sees another
instance already running, exposed at $XDG_RUNTIME_DIR/keyring/control.
After a 25s timeout, gnome-shell aborts the request without ever prompting
for a new password.
Because it is both problematic and pointless to query secrets in this case,
let's avoid it altogether and just prompt the user for the network password.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3646>
* Remove obsolete apps
* Add new core apps
* Deduplicate apps which were in both the grid and a folder
* Split system tools out of the utilities folder
* Rearrange the dash a little
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3642>
We currently create the default folder with the corresponding
app list, regardless of whether the apps are actually part of
the default install or not.
This matters when a user explicitly install such an app later,
as it will be hidden away in the folder rather than appended
to the app grid as expected.
To avoid that, only add currently-installed apps to the folder
when creating it.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3632>
Defining default apps as serialized GVariants isn't very human-friendly,
which likely contributes to the fact that the lists are in parts horribly
outdated (Books! Cheese! Screenshot! gedit!).
Instead, generate the lists at build time from simple text files, which
should be much easier to update.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3632>
When there are 3 or more windows in WorkspaceLayout, showing or hiding
window preview overlay in some certain orders could cause
inconsistencies in windows' vertical arrangement.
Let's take window A, B, C as an example. Initially, A is above B and B
is above C in workspace layout like this: A -> B -> C.
After opening activities, user could:
1. Move cursor to B preview, which would move B above all in layout:
B -> A -> C
2. Move cursor from B to C preview. When C's showOverlay() is called
before B's hideOverlay(), _restack() would move C above all and don't
change B's arrangement:
C -> B -> A
3. Finally, move cursor away from C's preview:
B -> A -> C
In this case, when user closes Activities, they would see window
stacking wrong for a while.
This commit adds some extra logic in _restack, checking the
_stackAbove's _stackAbove when this._stackAbove._overlayShown is true.
Though it's still not guaranteed to be always consistent as there could
be several WindowPreview with _overlayShown as true if pointer moves
really fast, this helps avoid glitches in many cases.
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4638.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3460>
Using the user object at `/org/freedesktop/login1/User/self` is
convenient, but has the caveat that login1 does not emit the
`PropertiesChanged` signal for the object.
That is indeed logical, as for signal emissions there is no
sender that can be used to resolve `self`.
The new TimeLimitsManager depends on change notifications for
user properties, so stop using the `self` shorthand and instead
create the User proxy for the user's UID.
Related: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/8185
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3636>
Most of the function is already asynchronous, except for the
initialization of the returned proxy. gjs' D-Bus wrapper gained
some convenience API a while ago that makes this trivial enough,
so use it.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3636>
Traditionally, getter/setter functions have been considered a
C convenience, and we therefore didn't bother to add them for
many properties that are only consumed from JS.
However now that gjs optimizes property accesses by calling the
appropriate getter/setter instead, it makes sense to add them.
Leave out ScrollView's [vh]scrollbar-policy properties, as they
have a combined `set_policy()` setter that is consistent with
GTK's ScrolledWindow.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3620>
Some properties have getters and setters that don't follow the
usual naming scheme, and they are therefore not recognized
automatically. Annotate the properties so that g-ir-scanner
picks up the methods.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3620>
Traditionally, getter/setter functions have been considered a
C convenience, and we therefore didn't bother to add them for
many properties that are only consumed from JS.
However now that gjs optimizes property accesses by calling the
appropriate getter/setter instead, it makes sense to add them.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3620>
