Instead of building the image on a schedule, use a user-incrementable
tag and try to build the image on each pipeline.
If the image tag does not exist, a new one is built and pushed.
If its already built the job is a no-op and the existing image
is used.
They were originally included was to educate users about what the icons
mean, for when they appeared in the top bar. However, since we no
longer plan on showing the status icon in the top bar, it's not so
important that people learn the meaning of the icons.
See https://gitlab.gnome.org/GNOME/gnome-control-center/-/issues/1421
...on the title bar with joined toggles (like in Background panel) for a
clearer UI.
Note 2020-08-26: This patch has been in Endless OS since 2014. It seems
like a good UX change, so I'm proposing it upstream with some minor
changes. - mwleeds
In the ubuntu 20.04, we met an issue about the output volume-slider
on the machine with the legacy HDA audio driver, the output device
is the Speaker first (analog-stereo pa sink), then we connect a hdmi
monitor, the HDMI audio is in the output combo-box, we select the
hdmi audio (hdmi-stereo pa sink) from the combo-box, the hdmi audio
becomes the active output device now, we adjust the output volume from
the volume-slider, the slider UI is changed, but the output sound
is not changed with the UI.
The root cause is when the speaker is active, the pulseaudio only
keeps the analog-stereo sink, the sink hdmi-stereo is unlinked, when
users select the hdmi audio from UI, the pulseaudio will unlink
analo-stereo sink and create hdmi-stereo sink, but before hdmi-stereo
is created, the output_device_changed_cb() is called and
gvc_mixer_control_get_stream_from_device() returns a NULL since the
hdmi-stereo sink is not created yet in the pulseaudio. Because stream
is NULL, the output_volume_slider->stream is NULL, users can't change
the output volume via the volume-slider.
To fix it, we add a output_volume_slider->stream check in the
device_update_cb(), if it is NULL, get the stream and set it to
volume-slider. In this function, the sink hdmi-stereo is created
already, so the stream is not NULL. And this change also applies to
input as well.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
g_time_zone_new() is deprecated in favour of g_time_zone_new_identifier().
g_time_zone_new_identifier() returns NULL if no matching identifier is
found, so warn and fallback to UTC in that case.
This commit changes the switch to do a proper delayed state change using
the state-set signal. Also changed is that we always update the switch
state rather than avoiding an update when it is not powered.
Avoiding this update was introduced in commit 4a009da483 (bluetooth:
Don't change the switch status when transitioning), however, the current
implementation causes us to get stuck in the wrong visual state
sometimes. Also, with this patch I am unable to see any visual glitch on
hardware that should be affected, and even if there was a glitch,
getting the final state right is more important.
Closes: #607, #1272
Rebuilding the UI would trigger the scale to be re-applied. This in turn
would cause the monitor to be snapped and possibly changing a correct
configuration.
Note that this really is a bug in the snapping code. But that code is a
mess and not fireing the signal is a good thing either way.
Closes: #1412
From https://gitlab.gnome.org/GNOME/gnome-control-center/-/merge_requests/1020#note_1211728
"
They are out of place - fine-grained control over brightness isn't what the
power settings are for. As a result, they confuse the settings overall.
In both cases we have alternative controls, through keyboard hot keys or
through the shell's screen brightness slider.
"
It's pretty clear from their experiences on smartphones that our users
know that Wi-Fi and other wireless technologies use enough battery that
turning them off is a power saving move.
The switches and text were also pretty confusing as we would be turning
"off" the devices to turn "on" the power saving.
Samba 4.12 complains loudly about a perl yapp driver module. Some
fiddling with the PERL5LIB and PRL_MM_OPT was required in order
to get perl to look into the /app prefix after bundling the module.
A logic error meant that the power profiles info boxes container
would not be shown if there wasn't a "degraded" performance as well.
Spotted by Allan Day
