This allows neat tricks like first arranging the script `fakegdb`:
```sh
function ignore_gdb_arg() {
if [[ "$1" == "--quiet" ]] || [[ "$1" == "--args" ]]; then
return 1
else
return 0
fi
}
ignore_gdb_arg "$1" || shift
ignore_gdb_arg "$1" || shift
echo exec "$@"
```
then running
```sh
meson test -C build -v perf-basic --test-args '--wrap "gdb --args"' --gdb --gdb-path `which fakegdb`
```
To make it possible to run the perf test case with gdb running the
actual gnome-shell.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1349>
This test runs all of GNOME Shell using the headless backend inside a
mocked D-Bus environment. The basic test tests, well, basic things, like
the panel menu, the overview, showing the app grid view, as well as
going back to the session view.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1349>
If one wants to run tests the non-installed gnome-shell, that currently
fails as gnome-shell the executable attempts to link against
./build/src/libgnome-shell.so, but when GObject introspection tries to
find what library to link to for Shell, it goes to the installed
libgnome-shell.so, causing two different versions of libgnome-shell.so
to be loaded.
This, however, can be avoided thanks to meson adding $ORIGIN paths to
relevant libraries before installing an executable. What this means in
practice is that we can inspect ourself upon startup, discover whether
the RPATH/RUNPATH header contains $ORIGIN, and if so, expand it to the
directory containing the executable, and prepend the introspection
search paths with said directory.
This effectively means that the introspection machinery now finds the
same library that the linker linked the gnome-shell executable with,
making it run successfully.
It's not possible to use $GI_TYPELIB_PATH since
g_irepository_prepend_library_path() takes precedence. There is no
"append" variant of that API.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1349>
Tearing down gjs means that we won't have any dangling references kept
alive by GC or otherwise alive Javascript objects, when we finally tear
down the mutter context. This allows for a clean shutdown of GNOME
Shell.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1349>
Commit 9168f6055e marked the parameter as nullable, which is in
conflict with the precondition check. But given that NULL is
valid for the underlying ClutterText (and GtkLabel accepts it
as well), there's no strong argument for disallowing NULL, so
remove the precondition check.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2465>
Only when part of it is actually visible.
Because the central rectangle is generally the largest part, this
eliminates most of the shadow's render time. For example, animating
`.workspace-background` by tapping Super, the shell's overall render
time is reduced about 15%.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1915>
WorkspaceBackground's allocate is a hot function called every frame
during overview animations. Port it to C.
While we're at it, cache the work area and the monitor geometry, which
do not need to be re-fetched on every allocation.
This reduces the average WorkspaceBackground allocation time from
0.134 ms to 0.017 ms. With four workspaces, scrolling the overview sees
an average WorkspacesView allocation time improvement from
1.104 ms to 0.678 ms.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2394>
The iterator was pointing to NULL when going out of scope, leading to
autofree not clearing the list.
==300183== 32 bytes in 2 blocks are definitely lost in loss record 14,798 of 38,939
==300183== at 0x484586F: malloc (vg_replace_malloc.c:381)
==300183== by 0x4D7D980: g_malloc (gmem.c:127)
==300183== by 0x4D95AB3: g_slice_alloc (gslice.c:1074)
==300183== by 0x4D96D96: g_slist_prepend (gslist.c:282)
==300183== by 0x487854F: shell_app_get_windows (shell-app.c:794)
==300183== by 0x48791A1: shell_app_get_pids (shell-app.c:1201)
==300183== by 0x488B293: shell_window_tracker_get_app_from_pid (shell-window-tracker.c:702)
==300183== by 0x488B632: get_app_from_window_pid (shell-window-tracker.c:370)
==300183== by 0x488B632: get_app_for_window (shell-window-tracker.c:436)
==300183== by 0x488B632: track_window (shell-window-tracker.c:549)
==300183== by 0x4CDBB75: g_cclosure_marshal_VOID__OBJECTv (gmarshal.c:1910)
==300183== by 0x4CD8BE9: _g_closure_invoke_va (gclosure.c:893)
==300183== by 0x4CF2A28: g_signal_emit_valist (gsignal.c:3406)
==300183== by 0x4CF2C2C: g_signal_emit (gsignal.c:3553)
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2367>
Even though GNOME Shell is not explicitly using structured logging via
G_LOG_USE_STRUCTURED, GLib uses it as default since 2016 [1], and so
we're de facto using it.
As per this, if backtrace on warnings is enabled, it is ignored since the
log handler isn't used anymore, and no dump is printed.
Thus, replace the default log handlers with writer functions instead, honoring
backtrace-warnings debug string.
[1] https://gitlab.gnome.org/GNOME/glib/-/commit/fce7cfaf40b6e1e50c9140aa0397f5
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/697>
Applications launched using D-Bus activation will have a returned PID of
0, which systemd interprets as moving the requesting process to a new
scope, causing GNOME Shell to be moved to another scope. Fix this by not
creating a systemd scope when PID is 0.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2305>
GTK4 added a convenience property for icon-only buttons. While that
use case is not quite as common in the shell as in GTK apps, it still
seems common enough to mirror the GTK API.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2254>
We started tracking all windows to make sure the assumption that any
window can be match to an app holds. It is not expected however to
ever represent OR windows in the UI, so it seems better to exclude
them from get_windows() instead of expecting everyone to filter the
return value themselves.
(The returned list still includes "uninteresting" windows like attached
dialogs, which can be important for cases like the correct MRU order in
alt-tab)
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/5233
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2252>
When neither the theme node itself nor any of its parents specifies
an explicit text direction, we default to aligning to the left.
That's a good default for LTR locales, but for RTL aligning to the
right is a better one.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2247>
As first mentioned in commit 672171093, the CSS spec defines shadow
colors independently from the colors casting the shadow. It's not
a physical light simulation so a shadow is allowed to be a different
color from texture casting it.
This means we only care about the shape of the source where alpha
values of zero are adjacent to alpha values of non-zero. And all such
non-zero pixels should be treated as fully opaque for the purpose of
shadow generation. While this would be wrong for a physical light
simulation it does allow us to cast shadows around semi-translucent
shapes and better support CSS.
Fixes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4477
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1918>
ClutterEffects are responsible for queueing redraws when their
properties change (and StScrollViewFade is a good citizen already), also
Clutter itself should queue a redraw when adding/removing an effect.
Users of Clutter should never have to queue redraws themselves (unless
they're implementing a custom ClutterEffect or ClutterContent), so don't
queue a redraw here.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2189>
The fade effect can also be added to the scroll view programatically
instead of using CSS via the st_scroll_view_update_fade_effect() API.
We make use of this API in the appDisplay, but since commit ba547ec1d
the fade margins get overridden to 0.0 from the ::style-changed handler.
Fix this by only setting the fade margins when CSS actually defines a
custom vfade/hfade offset.
Related: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/5079
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2189>
Since now the resource consumption of the "style-changed" signal
handlers should be significantly lower, we can start emitting the signal
in all cases again as it was originally intended. This fixes some small
visual issues, like the battery percentage in the panel only being shown
on hover or updating the size of StIcons on scale-factor changes.
This effectively reverts f74c07b9ac
Fixes https://gitlab.gnome.org/GNOME/gnome-shell/issues/1708
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2099>
Just like with the last commit, listen to some ClutterText property
changes to catch style changes that were trigerred by
_st_set_text_from_style() and invalidate the shadow spec on changes.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1744>
We update a few properties of ClutterText when the CSS changes via
_st_set_text_from_style() (which we call when receiving the
style-changed signal).
Right now we simply invalidate the text shadow every time we receive the
style-changed signal, but we're going to change that with the next
commits. To ensure the shadow still gets invalidated on CSS changed that
might affect the shadow, listen to a bunch of property changes that will
get notified when any property that affects the shadow changes.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1744>
When text shadows are used, the cursor indicating the current position
also casts a shadow.
This means we have to regenerate the shadow texture after the cursor
position changed, so invalidate the shadow in that case.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1744>
It's slightly more efficient not having to do property lookups. While
that is unlikely to be a concern for the properties in question, it's
still good practice and makes the code base a bit more consistent.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2168>
We are now consistently calling notify() when a property does change.
With that we can opt out of g_object_set()'s implicit change notifications,
so that notify is only emitted when a property *actually* changes.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2168>
It's slightly more efficient not having to do property lookups. While
that is unlikely to be a concern for the properties in question, it's
still good practice and makes the code base a bit more consistent.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2168>
- move all icons to the icons directory
- rename some icons to be more meaningful
- put all icons on a resource sheet
- update references to icon name changes
- deprecate icons for those in standard set
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2141>
The calendar's timezone cache holds only timezones already received
or added to the calendar, thus when asking it for a timezone for "the first
time", it returns NULL and a wrong timezone is used instead.
The get_timezone() does not do any I/O when the timezone is already cached
on the client side, thus it's fine to use it.
This could exhibit with non-recurring events, which use custom time zones,
in which case the event is shown in a wrong time.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2145>
The cursor texture, scale and position is captured separately and
overlaid on top of the preview, and on top of the final screenshot
image. This allows toggling it on and off post-factum.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1954>
As mentioned in the last commit, we'll split up taking screenshots into
creating a GPU texture first, and later saving that to disk as a PNG.
For individual windows it's already easy to get a ClutterContent with
the texture using meta_window_actor_paint_to_content (), for the stage
it's not that easy and involves a few extra steps including X11 specific
ones.
So introduce a new ShellScreenshot API which does all that and provides
the caller with a ClutterContent of the stage:
shell_screenshot_screenshot_stage_to_content ()
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1954>
With the new screenshot UI we're introducing, we'll be capturing all
screenshots to textures on the GPU at first, and then create a PNG
stream from those textures at a later point. This will allow us to
present screenshots immediately to the user so they can inspect them and
select the right area before actually saving them to disk.
As a first step to make this work, introduce a new ShellScreenshot API
that writes an existing CoglTexture to a PNG output stream:
shell_screenshot_composite_to_stream ()
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1954>
A property for reversing the visible order of children is a bit odd.
It has also been unused by actual gnome-shell code since 2010, and the
somewhat related pack_start()/pack_end() API in GtkBox(Layout) is gone
in GTK4.
With that in mind, turn the property into a no-op and deprecate it,
so that it can be dropped next cycle.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2085>
Replace deprecated functions with their direct replacements:
- dep.get_pkgconfig_variable() → dep.get_variable()
- prg.path() → prg.full_path()
- source/build_root() → project_source/build_root()
In one case we need meson.global_source_root() that was only
added in meson 0.58, so bump the requirement to that.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2077>
Since commit 1807be1, we clear the fallback icon when a window is
removed, and notify the icon change. The notify call currently
happens after removing the window from the running state, but
before syncing the state (and possibly clear the running state
altogether).
That state is inconsistent and results in an assertion hit when
some code tries to re-fetch the icon in response to the notify
call.
Address this by updating the state before clearing the fallback
icon, so that the app will be in the correct STOPPED state after
removing the last window.
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4888
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2073>
We were not properly tracking the window used for the fallback icon.
This could trigger a crash, as disconnection of the signal handler might
happen on the wrong window, which in turn could cause the icon change
notification to happen on a destroyed ShellApp instance.
Fix this by tracking the window used for the fallback icon. Disconnect
the icon notify callback explicitly for this window only when it is
removed.
Also, just to be extra safe, make sure that the icon is never NULL even
if x11_window_create_fallback_gicon should return NULL for some reason.
Closes: #4436
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2065>
We currently track windows via MetaWorkspace's ::window-added and
::window-removed signals. Those aren't emitted for override-redirect
windows though, as those aren't actually located on any workspace
(see https://gitlab.gnome.org/GNOME/mutter/-/blob/gnome-41/src/core/window.c#L1322).
While that shouldn't be an issue as there's no good reason to look up
the ShellApp of an OR window, extensions can make modifications that
result in OR windows ending up in places that assume that every window
has an associated ShellApp.
We can either
- accept that extensions break stuff (including badly)
- carefully handle app-less windows everywhere
- extend tracking to OR windows
Opt for the last option, as that's the most user-friendly and least
disruptive one.
It's also simpler to track ::window-created and ::unmanaged, as we
don't have to track workspaces or windows moving between workspaces.
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4751
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2029>
The window-type property may change, and with it the skip-taskbar
property that decides whether we consider it "interesting".
As a result we can end up showing untracked window, i.e. one
which does not have an associated app.
Cover this case by simply tracking all windows regardless of
their type. This won't change the app's running state, as that
is solely based on the skip-taskbar property (which is enforced
for all previously excluded window types).
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4751
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2029>
The way it is currently calculated is broken for days with DST changes
or leap seconds and it is not needed anymore anyway. This will also make
the fix in the following commit simpler.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2023>
Windows from some applications, such as guake, are created as showing on
all workspaces. When these windows are put on the workspaces via
set_workspace_state() during construction, the first time the window is
added to a workspace in the loop triggers the shell app tracker which
then tries to move the window to its startup workspace. This makes the
window unsticky which triggers another set_workspace_state() which
tries to remove the window from all workspaces, but currently it is only
on the first one and then adds it to the startup workspace. Once that is
finished, the first set_workspace_state() continues adding the window
to the remaining workspaces, despite the window now no longer having
on_all_workspaces set to true.
When the window is now unmanaged, the window according to its internal
state is only found on the startup workspace, so it will only be removed
from that. This causes the assertion to fail that checks that the window
is no longer present on any workspace after this.
Fixes https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4720
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2014>
commit 8721c5db37 made StPasswordEntry
honor the 'disable-show-password' setting.
Unfortunately, it introduced a lifecycle bug where the signal handler
for noticing setting changes can out live the entry itself.
This commit fixes the problem by using g_signal_connect_object
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2004>
Icons that are changed while an actor is not mapped may not have a theme
node associated with, and thus we'd end up not updating them at all.
In fact we return early in st_update_icon(), and this was not an issue
until commit 0b1dfbf6f3, because we'd end up to update the icon anyways
once the style was changed (and so with a valid theme node), but since
such change we might not updating the icon if no theme detail changed.
To prevent this, add a flag to require an icon update when the theme
changed, if no successfully update happened earlier.
Fixes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/4568
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1983>
St's theme test now requires the libmutter-test library, which is only
built when tests are enabled. Instead of mandating a particular build
configuration in mutter, add a corresponding option in gnome-shell as
well.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1975>
