When an app folder is open and its panel is shown, there are
three main elements: the folder name, the list of apps, and the
Edit togglebutton. By default, the user can use TAB to jump
between the app icons and the Edit togglebutton, but when the
togglebutton is checked, the folder name becomes editable and
the focus jumps to it.
Unfortunately, if the user goes out of the editable name with
TAB, the focus only cycles between the Edit togglebutton and
the app icons, never landing over the editable folder name (which
remains editable until the Edit togglebutton is unchecked).
This patch fixed this, making the editable folder name focusable
when using TAB, as expected, when the Edit togglebutton is
checked.
appDisplay: Fix accessibility for application folders
Currently, application folders have a `checkbutton` role, which
is confusing for the user because it's not something that
changes its checked status, but a button that opens the contents
of the folder. Also, the only indication that this is a folder
and not a "normal" application is that role, which means that an
user must know that pushbuttons are applications and checkbuttons
are folders, which is confusing.
This patch fixes this by adding the "expandable" accessible
state to the folder button, thus notifying the user that it is
an element with extra elements inside.
It also adds the folder name and the type as the full container
label when opening a folder, thus ensuring that the user knows
where they are.
Florian Müllner [Thu, 5 Mar 2026 15:22:41 +0000 (16:22 +0100)]
appDisplay: Stop using toggle-mode with folder icons
This is essentially a left-over from the original folder design, dating
all the way back to commit 6e3e2d9 in 2013.
It no longer makes much sense with the current design iteration, where
we pretend that the icon pops out of the grid and becomes the dialog
(i.e. we hide the icon and zoom out the dialog from its position).
messageTray: Avoid notification double destruction after activation
The activation handler of a notification may indirectly have already
destroyed it before `activate()` continues and tries to destroy it.
This happened for example in the `FdoNotificationDaemon` activation
handler when no default actions are specified, which ends up calling
`destroyNonResidentNotifications()`.
Another case where this could happen is the `WindowAttentionHandler`
which activates the window and upon activation of the window destroys
the entire notification source, which also destroys the notification.
Florian Müllner [Thu, 12 Feb 2026 12:50:47 +0000 (13:50 +0100)]
layout: Wait for backgrounds to load, even when interrupted
We no longer get stuck when a background manager is destroyed during
the startup animation, but we now resolve before the new backgrounds
have been loaded.
Address this by keeping the returned promise around until it is fulfilled,
and only resolve it after backgrounds were loaded successfully.
Florian Müllner [Tue, 10 Feb 2026 14:30:46 +0000 (15:30 +0100)]
layout: Reject promise if background loading is interrupted
Currently if a background manager is destroyed while we are waiting
for the background to load, the corresponding promise never resolves.
To prevent this from happening (and the startup animation getting
stuck as a result), guard the loading operation with a cancellable
and reject the promise when it is cancelled.
Adrian Vovk [Fri, 27 Feb 2026 21:21:46 +0000 (16:21 -0500)]
systemActions: Call into gnome-session for suspend
gnome-shell previously talked to logind directly on its own. This,
however, will lead to code duplication: the various inhibitor handling
bugfixes I've made to gnome-session (i.e. handling the new return values
from CanPowerOff/CanReboot/CanSuspend) would have to be copied into
gnome-shell. So instead, this commit just moves the responsibility for
handling suspend into gnome-session.
Adrian Vovk [Tue, 20 Jan 2026 05:04:57 +0000 (00:04 -0500)]
systemActions: Separate CanReboot from CanShutdown
The reboot actions and shutdown actions are independent, and might have
independent availability. It's perfectly valid for a system
administrator to configure a system where one action is available and
the other isn't. This commit separates the availability of these
actions, and allows them to be independently shown/hidden
Adrian Vovk [Fri, 27 Feb 2026 20:58:17 +0000 (15:58 -0500)]
systemActions: Switch to updated CanShutdown() API
There's corresponding changes in gnome-session that report more detail
about the availability of the action. This allows gnome-shell to
appropriately hide the action whenever it is inhibited, and would lead
to a Polkit prompt.
Adrian Vovk [Tue, 20 Jan 2026 05:09:29 +0000 (00:09 -0500)]
endSessionDialog: Immediately return from Open()
There's no need to wait for the dialog to finish fading in before we can
return from Open().
The code, as it was before, would break dialog updates: the dialog was
already open, so the signal would never get emitted, so the method would
never return, and so gnome-session would remain blocked waiting for a
return value that will never come.
st/entry: Fix clutter action leak when unsetting icon
The name of the removed action did not match the added one anymore,
effectively "leaking" an action whenever an icon was shown and hidden.
This was primarily affecting the clear button in the shell search.
Fixes: 0bf49afda ("Replace ClutterClickAction with new ClutterClickGesture")
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/4095>
Sebastian Keller [Sat, 28 Feb 2026 21:26:15 +0000 (22:26 +0100)]
st/entry: Fix primary/secondary icon leak
_st_entry_set_icon() takes a ref when setting the primary/secondary
icon, but the ref is not removed when unsetting the icon or the entry is
destroyed.
Since clutter_actor_add_child() already adds a ref to the icon, there is
actually no need to manually add another ref. So stop doing that to fix
this leak.
This however also requires that StPasswordEntry takes ownership of its
own icon that it may set or unset multiple times. Since it is initially
floating we need to sink the ref for that.
Sebastian Keller [Fri, 27 Feb 2026 12:22:37 +0000 (13:22 +0100)]
workspaceAnimation: Assign movingWindow before creating window actors
The BaseWorkspaceGroup constructor calls _createWindows(), which uses
_shouldShowWindow() to determine which actors to create for the
animation.
In WorkspaceGroup _shouldShowWindow() this depends on a comparison with
this._movingWindow, which however only gets set after the
BaseWorkspaceGroup constructor has been called.
This was causing the moving window to no longer stick to the screen
when moved between workspaces.
Fixes: 96be47933 ("workspaceAnimation: Split out BaseWorkspaceGroup") Closes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/8992
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/4092>
Shivam [Sat, 21 Feb 2026 12:22:35 +0000 (17:52 +0530)]
appDisplay: Refresh folder preview when favorites change
Pinning an app from inside a folder did not refresh the
folder preview. FolderIcon._sync() was only listening
to folder GSettings changes, while pinning updates the
favorites setting instead.
As a result, _sync() was never called and the preview
could show stale icons or remain visible when the folder
became empty.
Fix this by also connecting _sync() to
AppFavorites::changed, ensuring the folder preview
updates correctly when favorites are modified.
Alessandro Bono [Fri, 20 Feb 2026 22:57:14 +0000 (23:57 +0100)]
environment: Use replaceAll()
The replace() function expects two parameters, likely the third one
was ignored. Use replaceAll() to replace all the occurrences, not
just the first one.
Alessandro Bono [Fri, 20 Feb 2026 21:42:03 +0000 (22:42 +0100)]
objectManager: Remove unused function
Introduced with 4b450bab1164b9b1225702238e539201974a68b6, this function
has never been used. Furthermore it is wrong because objectPaths is not
indexed. Just drop it.
Sebastian Keller [Sat, 21 Feb 2026 11:32:26 +0000 (12:32 +0100)]
appDisplay: Consider admin defined folders before creating default ones
The system administrator might have defined some default folders system
wide via dconf default values, so checking for user defined values is
not enough. This restores the condition to also skip creating default
ones if the value is not empty. It still skips the creation of default
folders if the empty value was the result of a user action to allow
users deleting all folders.
Fixes: b0d396c82 ("appDisplay: Only create default folders on the first start") Closes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/9062
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/4087>
search: Do not check for class cancellable state after awaiting
In an async function, after awaiting, the cancelled state for the
class-tracked cancellable may not respect the actual state of the
cancellable that has been used to perform the call.
Cancellable reset API is deprecated for long time and it's quite broken.
In fact if a search provider receives a cancelled event may end up
checking if the cancellable has been cancelled, but this may not be true
anymore after a reset.
Use instead new cancellable instances so that once one is used we can
consider the previous one completed and we can forget about it.
search: Catch async function errors and log them unless cancelled
In case a search function fails, we are not providing clear errors, thus
trace the errors when calling async functions from sync ones and when
catching errors.
Note that at the moment is unlikely to receive any cancelled error here,
since the remote search providers calls handle the error already, but we
still use `logErrorUnlessCancelled` instead of `logError` to be more future
proof or to cover the case in which some extension may throw a cancelled
error.
Jordan Petridis [Mon, 16 Feb 2026 02:06:55 +0000 (04:06 +0200)]
status/volume: Check the stream before accessing it
stream-updated will also get emitted when the stream is unset, and trying
to access it then breaks.
In such case we don't need to update the css class of the indicator as it
will be hidden, and we will get another stream-updated before we show
it again anyway.
main: Call RegisterDisplay once UI has initialized
RegisterDisplay is called to signal that graphics have settled,
which allows GDM to terminate pending greeters and plymouth.
Previously, RegisterSession was used for both recording the login
and terminating greeters. This caused a race condition where the
new user shell was setting up while the greeter was tearing down,
leading to configuration conflicts during the asynchronous handoff.
By splitting into RegisterSession (called early to record login) and
RegisterDisplay (called after graphics settle), the greeter teardown
is delayed until the new session has completed its initial modeset.
Fixes: 32a01330 ("main: Register session with GDM on startup")
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/4057>
Florian Müllner [Wed, 4 Feb 2026 13:12:25 +0000 (14:12 +0100)]
swipeTracker: Handle invalid position when finding previous point
Commit 8e29babca8adc677 switched to `findLastIndex()` to find
the previous point, which inadvertently changed what value is
returned for invalid input.
Restore the old fallback value and log a stacktrace, so we can
hopefully track down what causes the invalid input.
Sebastian Keller [Wed, 21 Jan 2026 21:54:39 +0000 (22:54 +0100)]
main: Only call gjs_dumpstack() on warnings from main thread
The mozjs code called by gjs_dumpstack() does not appear to be thread
safe and was sometimes resulting in crashes when triggered by error
messages from threads other than the main thread. And even when not
crashing, such a backtrace can only be misleading since only the main
thread is executing JS code.
revisto [Mon, 22 Dec 2025 21:35:03 +0000 (01:05 +0330)]
status/volume: Snap to 100% when over-amplification is enabled
Sound may be distorted when overamplifying. It is therefore
useful to make it easier for users to select exactly 100%
when allow-volume-above-100-percent is enabled, as that is
the highest non-distorted value.
revisto [Mon, 22 Dec 2025 21:32:32 +0000 (01:02 +0330)]
slider: Add mark snapping support
Add API to add and clear marks, similar to GtkScale. The
slider value will snap to a mark when close to it, making
it easy for the user to position the slider exactly at
the mark's value.
The unsnapped value is still tracked internally, to allow
escaping snap zones when the scroll step is smaller than
the snap threshold.
This is not a good habit anywhere and especially in JS code, I assume
the rationale for this was to ensure that cached instances of the proxy
were cleared, but this did not work properly (as it happened on
finalize), and it would still require [1].
So let's just avoid this and assume that the getters just return what we
expect.
gdm/util: Clear the request if user verifier connection is closed
Similarly to previous commit, in case the user verifier connection is
closed we should just clear the user verifier instance, since starting
from this point is going to be just invalid.
gdm/loginDialog: Reset the greeter proxy on connection closed
If gdm fails to start the display session the greeter proxy we are using
gets disconnected, however since commit 56ea95f2 we are now preserving
the greeter instance, and this will eventually lead to using an instance
of a disconnected proxy.
This eventually implies that after a login failure, we would never get
the `::session-opened` signal that actually triggers the session start.
And thus, even the user has properly authenticated, gdm will stick to
the authentication page.
So, instead of reverting such commit (that isn't wrong per se), consume
the greeter proxy connection `::closed` signal and use it to
re-initialize the greeter instance.
If we use a regular idle watch, any session inhibitors can prevent the
watch from triggering, leading to the break manager to never consider
the break as having been completed until the inhibitor is gone. This
could lead to situations in which the system appears to be stuck with
a dimmed screen.
Also start the idle time when the idle watch is created, otherwise any
previous idle time would count towards it, causing the watch to trigger
early, leaving it inconsistent with the rest of the break manager state.
Florian Müllner [Fri, 31 Oct 2025 17:14:14 +0000 (18:14 +0100)]
js: Use one-shot idle/timeout functions
gjs now provides overrides for GLib's one-shot idle/timeout
functions, so use them to save a bit of code and clearly
indicate that a callback is only invoked once.
Florian Müllner [Fri, 31 Oct 2025 17:13:54 +0000 (18:13 +0100)]
extensions-app: Use one-shot idle/timeout functions
gjs now provides overrides for GLib's one-shot idle/timeout
functions, so use them to save a bit of code and clearly
indicate that a callback is only invoked once.
Jonas Ådahl [Fri, 3 Oct 2025 15:02:37 +0000 (17:02 +0200)]
status/keyboard: Adapt to external source of keyboard layout
This adapts to API changes that allow an external source (e.g. remote
desktop API user) to control the current keyboard layout. This is done
in two ways:
If the external keyboard source sets the keymap as "locked", the
indicator is fixed to the short name of the external source, and will
change the indicator label when the layout index is changed accordingly.
The builtin input sources are not selectable, and the keyboard selection
keybinding is disabled.
If the external keyboard source doesn't set the keymap as "locked",
whenever the current keymap is from an external source, it's indicated
as a non-sensitive but selected input source. Selecting another keyboard
layout overrides the one selected by the external entity.
gdm/util: Set minimum display time for authentication messages
Short messages like "PIN locked" were previously displayed for only
480ms (10 chars × 48ms), making them impossible to read. Ensure all
messages are shown for at least 2 seconds.
Ramez Medhat [Wed, 21 Jan 2026 15:28:22 +0000 (17:28 +0200)]
quickSettingsMenu: Add actor group to focus manager
This allows focus to menu item using the tab key to match the behaviour in other menus.
Closes https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/8904
Jan200101 [Sun, 25 Jan 2026 21:33:49 +0000 (22:33 +0100)]
shell-app: Improve discrete GPU detection
The original logic only took into account if a GPU is the "Default",
meaning the GPU used during boot, which was intended for Laptops with
Hybrid Graphics.
The new logic uses the new Discrete key to figure out which GPU is the
most appropriate to use in order of:
- the first non-default GPU if it is discrete
- the first discrete GPU if it exists
- the first non-default GPU
Milan Crha [Tue, 13 Jan 2026 16:49:27 +0000 (17:49 +0100)]
endSessionDialog: Remove unused UPGRADE_RESTART dialog data
The gnome-software D-Bus OfflineUpdates API does not distinguish
between offline updates and offline upgrades, thus the UPGRADE_RESTART
dialog data are not used and can be dropped.
Milan Crha [Tue, 13 Jan 2026 13:45:32 +0000 (14:45 +0100)]
endSessionDialog: Switch from PackageKit offline updates API to gnome-software
Let the endSessionDialog depend on gnome-software when deciding whether
any offline updates are scheduled and what to after applying them,
instead of using PackageKit, which is not included in distributions
using rpm-ostree and similar.
Another advantage is rather than implementing similar calls for various
package managers and systems in gnome-shell, these implementations should
go into the gnome-software, which should know about them already.
projects / gnome-shell.git / log