`adw_combo_row_get_selected_item()` does not transfer ownership, so we
shouldn't attempt to free the returned item. Otherwise use-after-free
can occur.

Fixes #3240


(cherry picked from commit 38a84d07)

Using `g_autoptr()` for multiple variables on the same line is awkward,
and in this case, also incomplete as we need to free the full GList.

Also, fix a GtkExpression leak, as setting it on a combo row does not
take ownership.


(cherry picked from commit 62feddf7)

- Remove the top margin from the "Turn off Airplane
Mode", as this is already handled by libadwaita
- Fix all status pages not being vertically centralized


(cherry picked from commit cf75db86)

(cherry picked from commit 65926c57)

(cherry picked from commit 0915fb1d)

Otherwis keyboard navigation is broken.

Fixes ecb4a9e3


(cherry picked from commit 650bbca0)

Based on Fedora 41. This way, if we need a new image for some reason, we
won't pull in potentially breaking rawhide changes.

Previously, if the user tried to change the system accent color via the
command line while Settings was open to the appearance page, and the new
color was to the left of the old color in the UI, Settings would
incorrectly set the system accent color back to the old color.

The notification from gsettings would call reload_accent_color_toggles,
which reads the settings and then walks the widgets left-to-right to
toggle the button on or off. If the new color was to the right of the
old color, this would trigger the old color's `toggled` signal (changing
the setting back to the old color) and then trigger the new color's
signal (changing the setting back to the new one). If the new color was
to the left of the old color, then the new color's `toggled` signal
fires first (doing nothing, since the setting already matches) and then
the old color's fires (changing the setting back to the old color)


(cherry picked from commit 8a322054)

Most often, CcMaskPaintable is used to follow Adwaita accent colors.
So, add a property to make this easier, so we can deduplicate and remove
code in CcSplitRow and CcIllustratedRow.

Should make future use easier too.

(cherry picked from commit 470a7448)

The default `gdk_paintable_default_get_current_image()` already does the
right thing (tested by somewhere displaying the result from
get_current_image()).

(cherry picked from commit 7fab119a)

This is very useful (and needed on larger scales) for SVGs, but to avoid
duplication we should pull it into CcMaskPaintable. Normally, scaling
happens inside GtkPicture, but because of the masking we have to do that
ourselves.

This is essentially a follow-up from
https://gitlab.gnome.org/GNOME/gnome-control-center/-/merge_requests/2847
to make this work easier for all future use of CcMaskPaintable.

For unscalable resources, we still set a GkdPaintable from it for
convenience, but don't use the parent GtkWidget then.

(cherry picked from commit e9f73c01)

And remove an unnecessary `gtk_media_stream_set_loop()`.

(cherry picked from commit 3e394804)

Now we can use the same method as in CcSplitRow, of retrieving the
CcMaskPaintable paintable instead of saving the GtkMediaStream.

(cherry picked from commit a7985578)

This reverts commit 2fdb4fd5.

In the next commit, we will use the CcMaskPaintable resource scaling
instead.

(cherry picked from commit 5fa0bf93)

The new unified way is much easier to use.

(cherry picked from commit bf206c27)

This makes the reload not trigger multiple (3) times for a mixed-DPI
setup. This uses the same mechanism as CcBackgroundPaintable to avoid
unwanted resource updates when initializing.

(cherry picked from commit a5facc17)

For some assets (e.g. VP9 webm files), it might be better not to use
alpha masking, but to use luminance masking instead. We can avoid using
"bad" GStreamer plugins then.

That does require switching between alpha or luminance masking mode. Do
so by simply testing if the resource filename extension is `.webm`.

(cherry picked from commit 15bafacb)

Loading VP9 with an alpha channel requires the `vp9alphadecodebin`
plugin, which is from the GStreamer Bad Plug-ins. These are not always
installed.

Instead, convert the alpha value of these videos to monochrome, and then
using the previous commit their masked value will be identical.

To make this work, we do need to ensure bt709 color space, otherwise we
would run into https://gitlab.gnome.org/GNOME/gtk/-/issues/7124, which
is a result of
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3479.

Closes #3173

(cherry picked from commit 8de3dd0f)

From settings-mockups, and converted to B&W using the conversion script.

(cherry picked from commit c05639c0)

The "Add Printer..." button label is already self
explanatory. Also, "All Connected and Network
Printers" seems like a misplaced tooltip, as
the dialog is about adding a printer, not showing
connected printers.

Closes https://gitlab.gnome.org/GNOME/gnome-control-center/-/issues/3267


(cherry picked from commit dd32ddff)

The animated assets got slight updates in settings-mockups at
Teams/Design/settings-mockups@8ef8dcbceb0be811fc036f53a3962459015d1925

Take those and use the BW conversion script to update them here too.

(cherry picked from commit 0d7120a4)

Colord profiles have to be "connected" for most operations on them.
Turns out that by default, the currently active profile is always
connected, but others aren't. So we get criticals setting such a
profile system-wide.

So, make sure a profile is connected before setting it system-wide.

(cherry picked from commit 7de79a5b)

Otherwise we also cannot test them for equality.

Closes #3264

(cherry picked from commit c96d4534)

The crosshair length is labelled with marks at 1/4, 1/2 and 3/4, calling
them fractions of the "screen". However, these marks are useless,
because the screen size would have to be 8192 pixels for them too make
sense.

Furthermore, these mark labels are so long that they prevent the
Settings window from working properly at its narrowest width. See #3131.

So, just get rid of those marks, and make the row similar to all other
scale rows in this page. That means changing to AdwActionRow and putting
it in the scale GtkSizeGroup.

(cherry picked from commit 43dd6357)

By doing that, the predefined avatars are distinct for a11y purposes.


(cherry picked from commit 0eeaabb1)

Update to upstream version '47.2'
with Debian dir d161e51050a640357d9fdd4195e55152b2ae8c6c

Tuned has become the new default in Fedora 41, replacing PPD. Tuned does
offer the same DBus interface with tuned-ppd. However, because tuned
supports much more profiles, the tuned-ppd api is slightly incompatible
as it can return an "unknown" active profile. Some discussion here:
https://github.com/redhat-performance/tuned/pull/669.

This crashes the Power panel. Instead, let's handle this by not doing
anything when we get that profile. This means no radio buttons will end
up getting selected, but that's better than showing an active profile
that isn't really active. Note that tuned-ppd currently does not
actively announce an ActiveProfile change if the change came from tuned,
but it might start to in the future, so

We do this by adding a new enum value, outside of the range of actual
profiles, as we don't want to handle it as a real profile anywhere. This
also removes the manual enum counting, there's no reason for that as
enums are guaranteed to start at 0 and increment by 1.

Note that tuned-ppd currently does not actively announce an
ActiveProfile change if the change came from tuned, but it might start
to in the future, so also reset all radio buttons if we get an unknown
profile, just in case someone has the Power panel open and changes the
active profile via tuned.

(cherry picked from commit 131c5591)

Noticed some unneeded code while creating the previous commit.

(cherry picked from commit fd62b901)

Aurelien Jarno authored Nov 16, 2024 and Aurelien Jarno committed Dec 07, 2024

The test-network-panel test currently takes around 60 seconds to execute
on the riscv64 build daemons, which matches the timeout. Therefore the
test intermittently pass or fail.

Use the --timeout-multiplier option of meson to avoid that.

Fixes wrong GPU names when g-c-c is launched using dGPU


(cherry picked from commit 2b111f06)

The labels are remarkably longer when translated to Spanish or German,
and otherwise the dialog does not fit a mobile screen size


(cherry picked from commit 1659b7ab)

On non-English locales. The labels are remarkably longer when
translated to Spanish or German. This makes at least the Details panel
fit a mobile-size screen


(cherry picked from commit 93b16780)

network-manager-applet Recommends this but not nm-connection-editor
which we use now instead

Gbp-Dch: Full

Our UI would only sync with its gsetting value at creation. Instead,
let's update the UI whenever the gsetting changes.

This can be tested with:
`gsettings set
org.gnome.desktop.notifications.application:/org/gnome/desktop/notifications/application/org-gnome-builder/
enable true`

Notice how the "Notifications" row updates/doesn't update.

Fixes #2103


(cherry picked from commit 436d41ae)

Volume-level source streams without a name were causing the Sound panel
to crash.

Let's make sure we only expose streams with a non-NULL name.

Fixes #3197


(cherry picked from commit 542419d9)

Closes: #1091092
Thanks: asciiwolf

(cherry picked from commit 4a4c58f9)

See https://handbook.gnome.org/maintainers/making-a-release.html

(cherry picked from commit 490adea2f5dcb68b1c33a2e9b61bf469a6d9a07f)

(cherry picked from commit eca398728875086d218b1d40b5817514189eb344)

We have now a release-build job for "dist".

(cherry picked from commit 37b5b6de90157fb2211d95a88c63cf2c7ecbf5ad)

"release-service-sandbox" was for testing.

(cherry picked from commit 54526a029176c1c694442265588ce0414b05c9bd)

(cherry picked from commit 91fa1846)

In the past it was a slightly transparent black dialog that was resized
according to the calibrated monitor, but resizing doesn't let us control
the positioning, meaning we can only fullscreening. Prepare for that by
making it a regular GtkWindow (dialogs are deprecated anyway), and add
our custom CSS to make it black with grey buttons, as it was before,
only now fully opaque.

(cherry picked from commit 2e41b808)

Closes: https://gitlab.gnome.org/GNOME/gnome-control-center/-/issues/948
(cherry picked from commit 06168bf8)

The calibration state is reset, but the second time the assistant is
run, the tree view signalling isn't hit, meaning the values selected up
front isn't forwarded to the calibration. The assistant page was also
still marked as "completed" even though no state had been set.

What this meant is that even though e.g. the device kind was appearing
to be selected to something, the actual calibration state was still the
default (unknown), which would later hit a soft assert and a error null
pointer dereference.

Fix this by resetting the assistant pages' completeness to incomplete,
and run the same code path for when the tree view options were selected
when an assistant page is prepared.

Closes: https://gitlab.gnome.org/GNOME/gnome-control-center/-/issues/1196
(cherry picked from commit d0a550b2)

Better to use a FIFO-typo behavior than the current LIFO, as otherwise
the visual results will start from the bottom, not from the top.

Closes #3275

(cherry picked from commit d143f100)

Update to upstream version '47.3'
with Debian dir babd9dc4afc8af2cd32354a99e1dd8ad9fb7d3c0

tagging package gnome-control-center version debian/1%47.3-1