| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
| |
The gtk modules xsetting may contain a ':' separated list of modules,
where the order is meaningless. In order to avoid tripping with the
unexpected order, split and sort it before doing our checks.
|
|
|
|
|
|
|
| |
The xsettings test is failing with:
/home/carlos/Source/gnome/gnome-settings-daemon/plugins/xsettings/test.py:84: PyGIDeprecationWarning: Using positional arguments with the GObject constructor has been deprecated. Please specify keyword(s) for "schema" or use a class specific constructor. See: https://wiki.gnome.org/PyGObject/InitializerDeprecations
self.settings_sound = Gio.Settings('org.gnome.desktop.sound')
|
|
|
|
| |
https://bugzilla.gnome.org/show_bug.cgi?id=797382
|
| |
|
|
|
|
|
|
|
| |
The shell stopped displaying it, so the setting is now always
false.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/624
|
|
|
|
|
|
|
|
| |
The directories that are monitored by gsd-housekeeping are likely to
contain files from other users that cannot be deleted. This will result
in permission denied errors that can be safely ignored.
Fixes #26
|
|
|
|
| |
(cherry picked from commit 91f7e36a63df696d0e2e9ecc99057c257250b122)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The user could slide the brightness on the control panel or press the
function key to make the screen brightness to 0.
However, If the users do not know about the keyboard buttons, they will
have a hard time turning the display back from the control panel if the
display is too dim or completely off.
For example, systems like ASUS UX550GE's brightness is in the range from
1 to 120000. The brightness level 1 has the backlight level which is so
dim and it's also practically unusable.
This patch makes it limiting to 1% brightness as the lower bound for
avoiding that usability issue. #54
Signed-off-by: Jian-Hong Pan <jian-hong@endlessm.com>
|
|
|
|
|
|
|
|
| |
The linear interpolation done for the color temperature would fail if
the value is exactly 0 or 1. Such values can be hit, so the we need to
accept them in normal operations.
Fixes #92
|
|
|
|
|
|
| |
Notifications traditionally use Title Case in their names.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is similar to the previous commit, but separated out because it
links the disk space notifications to Baobab, which is more disjoint as
an app — the previous commit linked other notifications to control
panel capplets, which are fairly strongly linked to the implementations
in gnome-settings-daemon.
https://wiki.gnome.org/Initiatives/GnomeGoals/NotificationSource
Signed-off-by: Philip Withnall <withnall@endlessm.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This allows the notifications to be tied back to an application (or, in
this case, control panel capplet) which is related to it; and allows the
user to specify per-app preferences for how the notifications are to be
displayed.
https://wiki.gnome.org/Initiatives/GnomeGoals/NotificationSource
Signed-off-by: Philip Withnall <withnall@endlessm.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Some bluetooth keyboards have only 5 rows of keys, so the top row has the:
1-9, 0, -, + keys. These not only double as function keys using Fn + key,
but also as media-keys, e.g. next/prev song, using Ctrl + Fn + key.
Unfortunately the firmware for these keyboards is somewhat buggy and when
sending media keys they send along the Ctrl modifier, so we receive e.g.
"Ctrl + next-song" this commit adds bindings for Ctrl + media-key for the
media-keycodes send by these keyboards, so that these work as expected.
|
|
|
|
|
|
| |
The default encoding for GIOChannel is UTF-8, but rfkill event is binary
data. If the value is invalid UTF-8, gsd-rfkill-manager will fail to
receive rfkill event.
|
|
|
|
|
|
|
|
| |
On some slow tablets 1 second is not enough for the power-button press
which wakes the tablet from suspend to get processed by g-s-d. This
causes the tablet to immediately resuspend when woken using the
power-button and on tablets the power-button often is the only way to
wake the device.
|
|
|
|
| |
(cherry picked from commit 70cdaca47e5db9e778b05c95f8e55e84ce08d450)
|
|
|
|
|
|
|
|
|
| |
We have always been leaking the GSource which is re-created every minute
to check the current night light state. Unfortunately, now that we are
correctly using the timerfd, we not only leak some memory but also the
file descriptor, causing major issues.
Fixes #91
|
|
|
|
|
|
|
| |
In the case that we have no backlight (for the screen) the DBus
interface will partially work. Querying the brightness will result in a
static value of -1. Trying to set the brightness in any way will instead
result in a DBus error.
|
|
|
|
|
| |
A simple test to check that gsd-power will notice an update to the
brightness.
|
|
|
|
|
| |
This adds a test for the step size and rounding including cases where
the backlight only has very few steps.
|
|
|
|
| |
This tests the event compression for backlight brightness updates.
|
|
|
|
|
| |
This replaces the special cases and temporary files hack used beforehand
and enables testing of the async logic in the backlight code.
|
|
|
|
|
|
|
|
|
|
|
| |
There is no need for the backlight helper to be capable of reading the
current brightness value or returning the maximum brightness. These are
now read directly by gsd-power and these features can be removed.
The only thing left for the helper to do is to ensure the user is not
trying to set a non-backlight brightness sysfs attribute.
https://bugzilla.gnome.org/show_bug.cgi?id=758413
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Writing the backlight value can take quite long, in particular when we
also get the overhead of starting up the helper process for writing.
Add code to handle the writing asynchronously and compress multiple set
actions into one write. The notification of the change only happens once
all of the queued up set operations have finished.
A trivial bugfix included is that the OSD display is again shown only on
the affected (internal) screen.
Note that a possible further improvement would be to just run the
backlight helper once and communicate with it through stdin or some
other methods, avoiding the overhead of starting it up all the time.
There are a few minor disadvantages with this approach though, see
comment #16 and following for more information.
https://bugzilla.gnome.org/show_bug.cgi?id=758413
|
|
|
|
|
|
|
| |
This improves the logic for closing notifications a bit either closing
the UPS notification or the device notification. This is not a big
improvement through, ideally we would keep track of the notifications on
a per-device basis instead.
|
|
|
|
|
|
|
|
|
|
| |
If a new device is added, then the warning level callback is called
unconditionally. Because of this, it could happen that an external
device (like a mouse) could cause the low battery mode in gsd-power to
be entered. In particular, this will result in the screen to be dimmed
very agressively.
Fixes issue #83
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The udev based device manager used for wayland environments was built
based on the state of the gudev option instead of the wayland option.
This resulted in an build error when building with `-Dwayland=false`
since the gdk/gdkwayland.h header file is not present on systems without
gdk-wayland-3.0.
This patch puts the compilation of the udev based device manager behind
the wayland option and the gdk-wayland-3.0 dependency
|
| |
| |
| |
| |
| | |
.set() allows to set about any value (strings, integers, ...), for
strictly boolean data, .set10() seems to be a better fit
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If we have timerfd support we will use it to wake up at the appropriate
time. There is a fallback non-timerfd path in which we wake up every
second.
The problem is that we had forgotten to ever check for timerfd support.
Add a check to the build system which enables this if timerfd_create()
exists.
Closes: #77
|
|/ |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
This commit was not intended for the master branch.
This reverts commit 451acbfb92089a28b75c7823ff723e3751f39167.
|
|
|
|
|
|
| |
This commit was not intended for the master branch.
This reverts commit 4cc26ccf24e8e987ea871fb07812ba849c8fda3c.
|
|
|
|
|
|
|
|
| |
This adds the required code to use GUdev to monitor for the existence of
the rfkill device. This is relevant as the rfkill device might not exist
initially if the rfkill module is not loaded during login.
Fixes issue #52
|
|
|
|
|
|
| |
The only benefit of doing so is that we can print information about not
finding any rfkill devices. This is not really helpful overall, so
simply remove the code.
|
| |
|
| |
|