unlockDialog: Defend against absence of blur effect
widget.get_effect(...) can return null while locking the screen, resulting
in a TypeError. In this situation the screen ends up black with a cursor
but never going to sleep, and moving the mouse brings the old screen
contents up but does not allow unlocking.
unlockDialog.js assumes that widget.get_effect will return non-null,
but other places such as getWindowDimmer in windowManager.js go out of
their way to be more careful.
[smcv: Add commit message, remove hard tabs, add missing semicolon]
The labels were previously aligned by using top-padding, which was
relative to font-size, and therefore didn't always align the text
correctly when font scaling was enabled.
Benjamin Berg [Thu, 20 Aug 2020 08:08:01 +0000 (10:08 +0200)]
data: Only unset environment when the service actually ran
ExecStopPost= is executed even if ExecCondition= fails. As such, we need
to add a guard as we would otherwise clear environment variables on X11
sessions (rendering them utterly useless).
Instead of reading a GSettings and building the app ids list
every single time FolderView needs to check for something,
cache it before redisplay and reuse this cached list everywhere.
When promisifying async operations in commit 764527c8c, the
finish function for read_line_async() was sneakily changed from
read_line_finish_utf8() to read_line_finish().
That is, the call returns a Uint8Array now that requires an
explicit conversion to string.
Florian Müllner [Thu, 13 Aug 2020 21:35:27 +0000 (23:35 +0200)]
cleanup: Remove old compatibility code
Since gjs moved to mozjs60, return values of int8_t arrays can
no longer be treated as strings. We originally made the conversion
conditional to keep working with the (then) stable gjs release.
That was two years ago and we require a more recent gjs nowadays,
so there's no good reason for keeping the old code path.
When the last item of an IconGridLayout page is removed,
the page itself is removed too. However, the indexes of
items of next pages are not updated, which mess up the
layout manager state.
Update the page index of the items at forward pages when
removing a page.
While the performance framework was originally written to collect
performance metrics, driving the shell by an automated script is
also useful to ensure that basic functionality is working.
Add such a basic test, initially checking top bar menus, notifications
and the overview.
Eventually it would be nice to separate the automatic scripting from
gathering performance metrics, but IMHO that can wait until we switch
from gjs' custom imports system to ES modules.
Florian Müllner [Tue, 4 Aug 2020 18:41:31 +0000 (20:41 +0200)]
perf-tool: Expose --x11 option
Running with the X11 backend is no longer as easy as not specifying
wayland, so expose mutter's --x11 option to allow enforcing the X11
backend for testing.
Florian Müllner [Mon, 3 Aug 2020 16:33:57 +0000 (18:33 +0200)]
environment: Mark transitions as "work"
global.run_at_leisure() is used from automated scripts to schedule
a callback when the shell is idle. However since we moved away from
Tweener, animations are no longer taken into account; fix this by
marking transitions as "work" if the convenience ease() functions
are used.
The original scripting framework was based on SpiderMonkey's
pre-standard generators, and was simply translated to the
corresponding standard syntax when updating it to work with
recent JS versions.
We can do even better by using the standard async/await pattern
instead of generators/yield.
Florian Müllner [Mon, 10 Aug 2020 23:34:03 +0000 (01:34 +0200)]
dbusServices/screencast: Quote filename in pipeline
Otherwise Gst fails to parse the pipeline string if the filename
contains spaces, as all words following the first are interpreted
as additional Gst elements.
endSessionDialog: default to not installing updates on low battery
If the user's battery power is low, we should not check the checkbox to
install updates by default. Rationale: if the user's battery is not low,
it's very unlikely to run out during a normal system upgrade. Low
battery is defined as any level below 30%, matching our battery status
indicator.
We'll also change the battery warning to only display when battery is
actually low. However, we will still always warn on battery for full
system upgrades, since these are expected to take a long time.
Future improvement: it would be nice to make the checkbox insensitive
when on low power. However, I don't think we currently have a proper
style for insensitive checkboxes. I was unable to make it look good.
Lastly, note that I did not test this on a laptop. I tested this by
mocking the return values of _isDischargingBattery() and
_isBatteryLow().
A side effect of removing the action buttons in favor of a regular
submenu is that we are a lot less constrained by size. So instead
of lumping "Restart" in with "Power Off", make it a separate menu
item.
endSessionDialog: Support updates in restart dialog
We will split off restart from the existing shutdown dialog, and
instead offer it as a separate menu item in the session submenu.
But before doing that, make sure that the existing restart dialog
exposes the same feature set as power off.
Andy Holmes [Wed, 20 May 2020 23:50:09 +0000 (16:50 -0700)]
St Documentation: add and improve documentation for public classes
Much of St is undocumented, aside from input/output arguments. This is
no doubt because a lot of it parallels Gtk closely, but is worth
improving since many new programmers are not familiar with Gtk.
Florian Müllner [Fri, 7 Aug 2020 19:35:35 +0000 (21:35 +0200)]
ci: Remove pot file test
A recent Docker image update broke the test, as xgettext now prints
the following warning:
warning: a fallback ITS rule file '/usr/share/gettext-0.21/its/metainfo.its'
is used; it may not be in sync with the upstream
That is completely unrelated to what the test is meant to catch and
could be fixed by adding appstream to the image, but considering that
- the test didn't actually catch the last template string regression
- we no longer allow template strings in files that include translatable
strings (and enforce that with a CI job)
- as of gettext 0.20.2, the template handling really really is fixed
(we'll see)
let's remove the test rather than piling up more stuff in the container
image.
Hans de Goede [Wed, 15 Aug 2018 13:03:56 +0000 (15:03 +0200)]
endSessionDialog: Support rebooting into the bootloader menu aka ("Boot Options")
This implements the "Alt" behavior for the "Reboot" button as outlined in
the design here: https://wiki.gnome.org/Design/OS/BootOptions
Note I've tried implemeting this with the AltSwitcher class from
js/ui/status/system.js first, but that puts the button in a St.Bin()
which causes the button to think it is the only button on the dialog
and makes it have rounded corners on both of its bottom corners.
Jonas Dreßler [Sun, 26 Jul 2020 11:26:20 +0000 (13:26 +0200)]
st/private: Multiply position in fb coordinates with resource scale
The framebuffer we use for rendering shadows is scaled by the resource
scale, that means we also need to offset coordinates when translating
them to the framebuffers coordinate system.
So far we forgot to do that when translating the framebuffer using the
position of the actor, which lead to small rendering bugs of
text-shadows for actors allocated at non-zero origins. To fix that,
simply multiply those positions with the actors resource scale.
Jonas Dreßler [Mon, 3 Aug 2020 19:47:42 +0000 (21:47 +0200)]
appDisplay: Return false in acceptDrop when not accepting the drop
Fix what is probably a copy-paste error and return false instead of a
CONTINUE DragMotionResult which is only meant for dragMotion events, not
drop events. This makes sure we don't create a folder when dropping an
app over the drag leeways of another icon.
status/network: Use D-Bus to launch Settings panels
For more obscure network configurations, we need to launch the
corresponding Settings panel with additional parameters, so we
cannot simply launch the .desktop file.
However we can do better than spawning a command line: Control center
exposes an application action we can use instead, so the process is
launched with the appropriate activation environment and startup
notification support.
This fixes an issue where the indicator can be out of sync until the
RfkillManager (used by it) properties change.
The problem is that multiple instances of the indicator will use
the same RfkillManager instance (getRfkillManager() returns a singleton)
that only guarantees to emit the changed signal in two scenarios:
when the D-Bus proxy connects and when the proxy properties change.
If by the time an indicator is instantiated the RfkillManager's D-Bus
proxy is already connected, that indicator would only sync its state
when the RfkillManager properties change.
Let's fix that by always syncing the state on construction - in the worst
case scenario the RfkillManager's D-Bus proxy won't have connected yet
and the indicator state will be temporarily out of sync but once it gets
connected the indicator will sync again with the correct state.
Benjamin Berg [Fri, 13 Dec 2019 18:07:16 +0000 (19:07 +0100)]
windowManager: Wait for X11 services using systemd
To do this, we now wait for the start/stop job to complete. We also have
two targets in gnome-session to ensure that everything is working as
expected.
In order to start the services, we simply request the
gnome-session-x11-services-ready.target unit, and wait for it to become
available. To stop, we use the gnome-session-x11-services.target unit
which should stop all services in a way that is entirely race free.
This requires both gnome-session and gnome-settings-daemon changes to
work (which are in the corresponding merge requests).
Benjamin Berg [Tue, 9 Jun 2020 11:01:39 +0000 (13:01 +0200)]
data: Create generic org.gnome.Shell.target
Move the GNOME shell service file adapation for x11/wayland into the
target/service files. This means that the session definition can simply
pull in org.gnome.Shell.target, without having to care about whether it
is starting an X11 or wayland session.
Note that this currently requires fork'ing to do the test. This will
however not be needed in the long term when ConditionEnvironment becomes
available (see https://github.com/systemd/systemd/pull/15817).
We technically do not need to use template units. But doing so means
that the unit can be translated to the app id more easily (though it is
not yet completely clear how this should look like in the long term).
Benjamin Berg [Thu, 13 Feb 2020 17:39:32 +0000 (18:39 +0100)]
data: Use org.gnome.Shell prefix for systemd units
In general we want to move towards using reverse domain names for
systemd units. Doing this also means we have a consistent name between
desktop file and systemd unit, allowing us to create a generator that
pulls in the unit as defined in the sessions RequiredComponents.
Jonas Ådahl [Thu, 23 Apr 2020 18:46:44 +0000 (20:46 +0200)]
Move screencasting into a separate service process
Move the screencasting into a separate D-Bus service process, using
PipeWire instead of Clutter API. The service is implemented in
Javascript using the dbusService.js helper, and implements the same API
as was done by screencast.js and the corresponding C code.
Jonas Ådahl [Wed, 22 Jul 2020 16:22:38 +0000 (18:22 +0200)]
dbusService: Queue shutdown check on startup
If something started the service, but crashed before managing to make a
method call, we'd end up with the service running indefinitely. Fix this
by queueing a shutdown check immediately on startup.
Jonas Ådahl [Tue, 21 Apr 2020 16:06:06 +0000 (18:06 +0200)]
status/remote-access: Visualize recordings as screen recording
If a remote access is marked as a recording, visualize it the same way
as a built in recording. Also don't stop it if there is an actual screen
sharing going on, so that one can use a plain "recording" while still
disabling what is an actual screen sharing.
projects / gnome-shell.git / log