We're storing in the texture cache images and scaled images appending
the scaling factor to the key. When a file changes the cache key
corresponding to that file is removed, but not the keys for the scaled
ones so that images in the cache are never reloaded.
This patch removes all keys from the cache related to the file that
changes, including those with the scaling factor.
A new set (hash table) was added to keep track of scale used to be able
to remove all possible images in the cache.
When the KEY is removed from the cache, we can look now in the scale set
for and each scale we also remove the key "KEY1.000000", "KEY2.000000",
etc.
Assuming that the number of used scales is small (I would typically
expect one or two), the overhead should be negligible.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/567
When removing a string from a settings list, we iterate over all
existing entries and copy all strings except the one that's being
removed to a new list, which is then written to GSettings.
However we currently always increment the index, so we end up with
a NULL entry in place of the removed entry, which is then interpreted
as the end of the list. In other words, we also remove all entries
that follow the removed string.
Fix this by looping over the list entries instead of the index, and
only increment the index for entries we copy.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1946
Instead of every individual StThemeNode. There are essentially two kinds
of theme nodes: Those we create for lookups, and those interned by the
theme context and used by StWidgets. Listening to the signal on the former
is pointless as they are short lived and not meant to be really used for
drawing. So it is only essential to track stylesheet changes in those we
intern for later use.
This change does precisely that, it lets the StThemeContext track the
stylesheet changes and let all known theme nodes reset their state for
it.
The internal array holding all connected handlers for this signal in glib
was about the biggest single allocation made in gnome-shell, as interned
theme nodes nodes are around the 4 to 5 digit numbers. This essentially
makes it disappear.
This however means that widgets that are explicitly set a theme through
st_widget_set_theme() don't get their theme node implicitly updated.
There's little reasons to use that API, so perhaps this is an acceptable
tradeoff.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/779
A StWidget could get its style from a) a theme set in the StThemeContext,
and b) directly through it's ::theme property. Generally, overriding CSS
through the latter cannot be recommended as it loses any connection with
the global theme (eg. the ones you get through selector specificity).
It sounds a bit too powerful and pervasive, there's no use for it in
gnome-shell and doesn't look like something that could be recommended on
extensions. So, just drop this piece of API.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/779
By now, all containers and layout managers except StBin (and its
subclasses) use the generic ClutterActor expand/align properties
to control how their children are laid out.
This is particularly confusing as two or the properties StBin uses
for layout - x-align and y-align - shadow the generic ClutterActor
ones, but work very differently: They use a different enum and
determine how the bin lays out its child, instead of how the bin
is laid out by its parent.
Address this by deprecating the StBin properties and using the same
generic ClutterActor properties as everyone else.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/803
As per GNOME/mutter!385 [1], the compositor is finalized an its pointer
cleared on display close.
However, since the shell reacts to such events instead of controlling them,
when the shell is stopping or restarting and its display closing, the shell
stage destroys its children after the display closing is finished and during
this process the focus is unset, causing focus_actor_changed() to be called
and thus calls to meta_stage_is_focused() which deferences the now NULL
compositor, leading to a crash on shutdown.
Since after this point we should just ignore any stage event, disconnect
from them all.
[1] https://gitlab.gnome.org/GNOME/mutter/merge_requests/385https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/746
If graphical applications want to start from systemd units, they need to
start after we're properly ready to display them. This is particularly
important under X where `_GTK_FRAME_EXTENTS` and other xprops are needed
to have the right theming.
We're doing this in an idle callback so that the dynamic starting of
`gnome-session-x11-service.target` (which launches `gsd-xsettings`) as
the result of a signal emission happens before us signalling we're ready
for later things to start.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/750
Until commit 506b75fc7f we got away with not handling a NULL return
value, as cairo_surface_destroy() deals with a NULL surface; the same
isn't true for get_width/get_height, so guard to code in question to
prevent a crash.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1678
This is now handled in Mutter. Also, respect the result size instead
of assuming it to be equal to the clip size, as the clip takes actor
coordinates while the result takes buffer coordinates.
This can be fixed in a future API iteration.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/728
We allow opening new windows as a fallback in case the app doesn't give
us explicit information about it, but we don't want to allow opening new
windows if we're unable to ask for this information (we can only use the
APIs to get this information while the app is RUNNING).
So always return FALSE in case the app is STARTING, always return TRUE
in case the app is STOPPED (starting an app always opens a new window)
and go through the usual checks in case the app is RUNNING (and
eventually fall back to FALSE).
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/121
The use of box-shadow on a StWidget that has a background-gradient was
not been rendered correctly, the shadow borders was calculated inside
the st_theme_node_prerender_shadow function and in the case that we've a
prerendered_texture the max_borders was not calculated and are 0.
This patch creates a new static function to compute shadow maximum
borders copying the code from st_theme_node_prerender_shadow, and call
this new method in the case that we've a prerendered_texture.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1186
Evolution draws in libsoup, which exposes deprecated types in its
API. While its headers have been fixed to guard the affected symbols,
those fixes aren't in our CI images yet.
Until we get a fixed version, just disable all deprecation warnings
during the include in order to not trip over "foreign" bugs.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/709
We aren't using any deprecated evolution-data-server API, so we can
turn it off; this avoids compiler warnings for glib deprecations
used by those functions, which makes it harder to spot warnings for
our own code base.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/709
We translate the raw stream content far too directly into a char*,
it notably forgets that the stream does not have nul-ended data,
this means we are potentially adding garbage after the pasted content.
Tentatively fixes https://gitlab.gnome.org/GNOME/gnome-shell/issues/1570
An Endless OS system was found in the wild with a malformed
.local/share/gnome-shell/notifications. When deserialized in Python,
after passing trusted=True to g_variant_new_from_bytes(), the first
element of the first struct in the array looks like this:
In [41]: _38.get_child_value(0).get_child_value(0)
Out[41]: GLib.Variant('s', '\Uffffffff\Uffffffff\Uffffffff\Uffffffff\Uffffffff')
When deserialised in GJS, we get:
gjs> v.get_child_value(0).get_child_value(0)
[object variant of type "s"]
gjs> v.get_child_value(0).get_child_value(0).get_string()
typein:43:1 malformed UTF-8 character sequence at offset 0
@typein:43:1
@<stdin>:1:34
While g_variant_new_from_bytes() doesn't have much to say about its
'trusted' parameter, g_variant_new_from_data() does:
> If data is trusted to be serialised data in normal form then trusted
> should be TRUE. This applies to serialised data created within this
> process or read from a trusted location on the disk (such as a file
> installed in /usr/lib alongside your application). You should set
> trusted to FALSE if data is read from the network, a file in the
> user's home directory, etc.
Persistent state is read from the user's home directory, so it should
not be trusted. With trusted=False, the string value above comes out as
"".
I don't have an explanation for how this file ended up being malformed.
I also don't have an explanation for when this started crashing: my
guess is that recent GJS became stricter about validating UTF-8 but I
could be wrong!
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1552
Instead of going via the MetaShapedTexture to get the cairo surface, get
it from the window actor. The window actor can then handle this in a way
that makes it include potential subsurfaces.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/692
Using the bus name to notify service startup to systemd has some
disadvantages. The main one being that systemd will consider a
gnome-shell restart (Alt+F2 r) a service failure and restart the shell,
cleaning up all its children (i.e. user launched applications). In the
future the shell should launch applications in their own transient unit
so that a service restart does not affect applications.
Another potential issue is that we must never load
gnome-shell-wayland.service and gnome-shell-x11.service at the same
time, as systemd does not like two services providing the same bus
name.
Fixes https://gitlab.gnome.org/GNOME/gnome-shell/issues/1496https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/690
Mutter's Clutter fork can no longer be initialized separatedly, as
its backend now draws from MetaBackend. Adjust the code to use the
newly added test initialization function instead to get the test
back up.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/691
Clutter's backend code depends on MetaBackend now, which makes it
impossible to initialize without resorting to private mutter API.
Luckily we only need Clutter for interactive tests which are broken
anyway, as Clutter.main() and friends were removed a while ago.
So for now, get at least unit tests working again by simply the
unnecessary Clutter initialization.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/691
Popups and other override-redirect windows are meaningless to everything
that depends on the ShellWindowTracker. Ignoring those windows will result
in less ShellApp::windows-changed signal emissions, and less activity in
the AppMenuButton and everything else that depends on them.
Reduces gnome-shell CPU activity while typing on the Epiphany addressbar,
as the pop up animation there results in a number of xdg_popup being
created and destroyed.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/642https://gitlab.gnome.org/GNOME/mutter/issues/556
Now that we allow to disable session mode extensions, it can be useful
to reset an extension to its original state, that is disabled in the
regular session, but possibly enabled via the session mode.
Add a corresponding command.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
Change both 'enabled-extensions' and 'disabled-extensions' keys as in
commit ce1bee727. While those actions are now also exposed by the
shell's D-Bus API, there is some value in allowing the tool to be used
outside a running GNOME session (for example in setup scripts), so
keep changing the GSettings keys directly.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
With the addition of the 'disabled-extensions' key in commit ce1bee727,
the way extensions are enabled/disabled changed: Now a UUID is always
added to one list and removed from another.
Prepare for that by generalizing the relevant bits of the existing
enable/disable commands as helper functions.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
The ability to install unaudited extensions directly from a zip file
can be useful for testing and code review, so implement a corresponding
command that complements the previously added 'pack' command.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
Extensions are uploaded to extensions.gnome.org as zip files that
not only contain the extension sources, but also compiled GSettings
schemas and message catalogues. To make this more convenient, add
a corresponding command for creating an archive suitable for up-
loading.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
We already support displaying extension details for the list command,
so it's a logical extension to also support showing extension info
for a particular extension (not least because the shell has a
corresponding D-Bus method).
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
In addition to a plain list of all extensions, add options to display
additional details of each extensions and to filter the list by
enabled state or install location.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
This implements more functionality of the existing tool and, as
'reload' is an unreliable feature that doesn't work more often
than not, the last bit that we will replicate.
The command follows the original for the most part, with the most
important difference being the installed template, which doesn't
provide any sample functionality and uses modern JS syntax.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
This replicates the most basic functionality of the existing
gnome-shell-extension-tool, albeit using a git/gio/gsettings
style command interface rather than plain options; this will
allow us to implement more complex commands that have options
on their own in the future.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1234
Whenever an app is installed, the usual routine is
to run 'gtk-update-icon-cache' after installing all
of the app's files.
The side effect of that is that the .desktop file of
the application is installed before the icon theme
is updated. By the time GAppInfoMonitor emits the
'changed' signal, the icon theme is not yet updated,
leading to StIcon use the fallback icon.
Under some circumstances (e.g. on very slow spinning
disks) the app icon is never actually loaded, and we
see the fallback icon forever.
Monitor the icon theme for changes when an app is
installed. Try as many as 6 times before giving up
on detecting an icon theme update.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/661
The texture cache, right now, only monitors for
complete theme changes. If the contents of the
icon theme change, however, the texture cache
isn't properly invalidated.
This manifests itself as a randomly reproducible
bug when installing an app; the app icon may be
the fallback forever, or as long as something else
updates the icon theme.
Watch for the GtkIconTheme:changed signal, and
evict the texture cache when the theme changes.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/661
StAdjustment implements the ClutterAnimatable interface, so we can
already animate its properties with ClutterPropertyTransitions.
But as it is currently not possible to associate a transition with
an adjustment, it must be owned (and kept alive in case of GC) by
the calling code.
Change that by implementing the same (add|remove|get)_transition() API
as ClutterActor, so we can use a familiar API and even duck typing in
case of javascript.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/669
We now have everything in place to replace Tweener for all animatable
properties with implicit animations, which has the following benefits:
- they run entirely in C, while Tweener requires context switches
to JS each frame
- they are more reliable, as Tweener only detects when an animation
is overwritten with another Tween, while Clutter considers any
property change
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/22
We currently only consider a remote "app.new-window" action when running,
but not a fixed "new-window" action in the .desktop file. The latter is
clearly useful as well, in particular as open_new_window() already does,
so add it.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/665
`g_object_notify()` actually takes a global lock to look up the property
by its name, which means there is a performance hit (albeit tiny) every
time this function is called. For this reason, always try to use
`g_object_notify_by_pspec()` instead.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/652
In `st`, we can do this by using `ST_PARAM_*`. In the other code files,
just use `G_PARAM_STATIC_STRINGS` directly.
This is just a minor convenience to prevent a few unnecessary string
copies.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/646
A generic, introspectable Shader effect is not only more flexible
than a shader actor, it will also make it much easier to turn
Lightbox into an actor subclass and replace Tweener with Clutter's
own animation support.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/651
Since -Werror=missing-braces is used, having missing braces warnings
aren't allowed. However, the first member of struct sigaction is a union
whose first member is a pointer, causing clang to produce warnings when
it is initialized to { 0 }.
Instead of initializing to a zero value, we can specify values of
members directly in the initializer to avoid warnings.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/633
Glib stopped providing any fallback implementations on systems without
memmove() all the way back in 2013. Since then, the symbol is a simple
macro around memmove(); use that function directly now that glib added
a deprecation warning.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/632
st_theme_node_paint_equal() was originally added to preserve paint state
when a style change didn't affect any of StWidget's cached background
resources.
That's why using it for filtering out unneeded style changes as in commit
f662864a misses any non-background related properties that are relevant
for subclasses. Add additional tests to make sure we keep emitting the
signal in those cases.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1212
This is a small convenience wrapper around clutter_color_equal()
for the different color components, which also handles the case
where one (or both) of the icon colors are %NULL.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1212
Since priv->device gets set to NULL inside st_button_release, ungrab the
input device before calling st_button_release and avoid
clutter_input_device_ungrab failing with a critical error.
This fixes a regression introduced with
d5a1a888d9
While at it, also remove the superfluous line resetting priv->device to
NULL and move the check for priv->grabbed into an elseif block since
there should be no case where StButton has both grabs at the same time.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/614
Since the removal of the old (pre-3.16) message tray, legacy tray icons
are very unlikely to be placed in a container that is animated using the
deprecated anchor point.
Just assume that the regular stage position is good enough.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/572
If the state we're trying to delete does not exist, do not log an
error.
Prevents this journal warning at startup:
gnome-shell[1082]: Could not delete runtime/persistent state file: Error removing file /run/user/1000/gnome-shell/runtime-state-LE.:0/screenShield.locked: No such file or directory
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/555
Instead of considering a GValue containing a NULL string to be a
programmer error, simply return NULL.
remove_mnemonics() is in fact called on the value of the
"choice-label" property as well, which has NULL as its default
value.
This prevents triggering the following gnome-shell warning:
gnome-shell[1082]: remove_mnemonics: assertion 'label != NULL' failed
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/555
Util.ensureActorVisibleInScrollView takes care of the potential scroll view fade
effect in order to compute the scroll offset, reading the ScrollViewFade's
`vfade-offset` property. This was correctly working until gnome 3.30 cycle.
However such property isn't defined now because since gjs 1.54, it can only
fetch introspected properties and St.ScrollViewFade was considered a private API
not exposed by gir.
Fix this by also introspecting st-scroll-view-fade sources.
Not being considered private anymore, install the header.
Fixes https://gitlab.gnome.org/GNOME/gnome-shell/issues/1061
Updating the :first/:last-child pseudo classes can result in a lot
of unnecessary style changes when bulk-adding children to a container,
as every child ends up as the new last child.
Address this by deferring the style change to an idle, so we only do
the work once for the actual first and last child.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/529
Images are loaded either with a supplied fixed size, or using the "native"
dimensions of the file. When creating a content image from the loaded data,
we currently simply apply this directly to the preferred size.
This works usually fine: GdkPixbuf will always keep the aspect ratio, so
if only one dimension is provided, the other will be adjusted accordingly:
Loading a 200x200 image with a requested size of (100, -1) will result in
a 100x100 content image.
There is a catch though: GdkPixbuf will only scale *down* to the requested
size, no up. That is, loading a 100x100 image with a requested size of
(200, -1) will result in a 100x100 pixbuf. But as we assume that the pixbuf
size matches the requested size, the image content ends up with 200x100.
Fix this by explicitly handling the case where only one size was supplied,
and make the other dimension take the aspect ratio into account
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/525
The parameters that may affect the icon on ::style-changed are more size
related than visual (we listen to icon theme changes for the latter). It
makes sense to just update the icon if the size came out different.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/524
With a high enough amount of actors, there may be enough theme nodes and
signal connections on StTheme::custom-stylesheets-changed that
g_signal_handlers_disconnect_by*() on dispose becomes expensive, this may
become a surprisingly hot spot in StWidget::style-changed.
Keep the handler ID around and use g_signal_handler_disconnect() to avoid
linear lookups for the matching func/data.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/524
A window being unmanaged can cause the ShellApp to be removed from
the ShellAppSystem, which if we are unlucky is the app's last
reference, causing it to be disposed and freed. It would be bad if this
happened before we finished handling the signal.
Use g_signal_connect_object to ensure that a reference is held to
the ShellApp for the duration of the signal handler, delaying its
last-unref.
In particular, when a signal handler calls _shell_app_remove_window(),
there is a brief period for which ShellApp breaks the intended
invariant (see !497) that app->running_state is non-NULL if and only if
app->running_state->windows is also non-NULL (non-empty). Freeing the
ShellApp at this point would cause a crash. This seems likely to be the
root cause of <https://gitlab.gnome.org/GNOME/gnome-shell/issues/750>,
<https://gitlab.gnome.org/GNOME/gnome-shell/issues/822> and
<https://bugs.debian.org/926212>.
Signed-off-by: Simon McVittie <smcv@debian.org>
Compare painting/geometry of old and new paint nodes, so it's ensured to
be only emitted on actual style changes. Emission still must be propagated
through to children, though.
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/issues/1153
Actors themed through CSS should ideally get sizes and positions that
conform to the "pixel grid". A notorious example is the panel that has a
height of 1.86em. On unchanged font settings and hidpi that translates to
55px, which leaves the workarea with "half pixels" that hidpi wayland
applications don't know how to fully cover.
If the requested height is a multiple of the scale factor, the workarea
and maximized applications can then work on full pixels.
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/issues/91
The same code for reading the current magnifier state is repeated in both
shell-recorder, shell-screenshot and magnifier itself.
So to move this inside a property of st-settings so that we can refer to it
all over the places removing duplications.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/473
If the locale doesn't honor strftime()'s %c argument, we would end up
feeding NULLs into GdkPixbuf tagging. Fallback to a sensible (although
not nicely localized) datetime string.
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/issues/1017
Just like we take a remote "new-window" action into account for
opening new windows, we should call an explicit "quit" action
before falling back to closing all the app's windows on quit.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/624
For window backed apps, create_icon_texture() doesn't return an StIcon
but a generic widget. Set an appropriate style class to make it easier
to apply a specific style only to fallback icons.
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1027
With StImageContent, the meaning of passing -1 as size parameter changed
from "load the image at its preferred size" to "abort the session". It
is therefore no longer possible to just load the image and then have it
scaled by applying a CSS size to the texture's parent.
Setting the size from CSS is useful though, so to still allow that, fall
back to the actor's size (which can be determined by the style).
https://gitlab.gnome.org/GNOME/gnome-shell/issues/1027
We need this to run `test-theme`, otherwise when run as part of the
build tests it fails like:
error while loading shared libraries: libmutter-cogl-4.so.0: cannot
open shared object file: No such file or directory
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/442
The executable is assumed to be run from $top_srcdir/src, which is
essentially an autotools left-over (it's where the program ended
up with srcdir == builddir).
Now with meson, its actual srcdir makes more senses.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/419
Clutter no longer hard-codes a resolution of 96 DPI (although that's
still the default), so any assertions of sizes for physical units
may be off.
Fix this by setting up the test environment according to the
assumptions.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/419
Since we started to support tray icons on wayland, the icon we show
is not the actual XEmbed window. Instead, we let mutter create a
MetaWindow for it, then use its window actor as source for a clone
we (or nowadays: extensions) can add, remove and destroy freely.
To not let the real icon get in the way, we set an empty input shape
and make its window actor fully transparent. This works OK on X11,
but on wayland all events still go through Clutter, so any reactive
surface actor inside the window actor will block events for any actors
underneath (and status icons go into the top-windows group, so almost
all actors are affected).
Luckily we can pile another hack onto the pile of status icon hacks ...
https://gitlab.gnome.org/GNOME/gnome-shell/issues/191
While mapping the :first/:last-child pseudo classes directly to the
ClutterActor:first-child/:last-child properties allows for an easy
implementation, it is unexpected that rules can appear to not have
an effect because the selected child is hidden. GTK's behavior of
applying the classes to visible children makes much more sense, so
change our implementation to do the same.
https://gitlab.gnome.org/GNOME/gnome-shell/merge_requests/312
