The blur effect needs to take the scale-factor into account, so we
listen for scale changes. However we set up the signal handler when
creating a background, which is repeated for each monitor, and every
time the monitor configuration changes. But we only disconnect the
last handler that was connected, and only when we are destroyed,
not when recreating backgrounds.
Fix this by splitting out updating the effect parameters to a separate
method that iterates over all backgrounds, so we can simply set up the
handler from the constructor.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1341
getEvents() filters all events for the given range and sorts the result.
That's more than we need when checking whether there are any events,
where we only care that there's at least one event in the range.
Address this by splitting out the event filtering into a generator
function, so hasEvents() can return after at most one iteration.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1192
We track messages so that we can account for just added and removed
events instead of having to rebuild the entire list, however it's
also possible that the time or summary of an existing event changed.
Account for that by updating existing messages in-place.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1192
The performance of the icon grid was being hindered by a large number
of primitives (a few hundred) being copied from the CPU to the GPU on
each frame. This was first noticed in mutter#971 but we failed to
investigate all the issues at the time.
You can also see the high number using `COGL_DEBUG=batching` or
`COGL_DEBUG=disable-texturing`. So now it's obvious that high number is
every letter of every label being uploaded as a separate quad. Let's not
do that and instead treat the whole label as a single quad/texture.
Measured performance on an i7-7700 at UHD 3840x2160:
Journal entries per frame on the icon grid:
* Before: 288 (18 KB copied from CPU to GPU)
* After: 73 ( 4 KB copied from CPU to GPU)
Spring animation:
* Before: 20-30 FPS, avg 22/peak 45 milliseconds per frame
* After: 30-40 FPS, avg 14/peak 28 milliseconds per frame
Scrolling the icon grid:
* Before: 15 FPS, 50 milliseconds per frame
* After: 30 FPS, 28 milliseconds per frame
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1329
(cherry picked from commit ae338af1e8)
Commit c7e597cf72 tried to improve the slide animations when entering
the overview by using the same time as the overall overview animation,
but in fact broke the animation most of the times.
That is because the Overview imports OverviewControls before defining
the ANIMATION_TIME variable, so any javascript code that is evaluated
during that import will see the value as "undefined" (which is converted
to 0 for the animation).
Fix this by moving the ANIMATION_TIME variable before the imports instead
of the usual placement.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1331
(cherry picked from commit 26d27fdbf8)
When you tap Super and see the sidebars and windows slide, it looks more
cohesive if those animations complete at the same time.
Previously there were 0.09 seconds difference between the two animations
which was enough to make it look slightly buggy. Now it doesn't.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1289
(cherry picked from commit c7e597cf72)
Animating the window clones of the overview requires the fullGeometry
and the actualGeometry to be set, which they won't be when showing the
overview for the first time. So don't even try to animate the window
clones in that case because the geometries will still be null and
accessing them in workspace.js will throw errors.
The workspace views will still get the correct layout as soon as the
allocations happen because syncing the geometries will trigger updating
the window positions. Since animations are disabled for position changes
when syncing the geometry though, we won't get an animation and the
clones will jump into place. That's not a regression though since before
this change we also didn't animate in that case because the geometries
used were simply wrong (the actualGeometry was 0-sized as explained in
the last commit).
If we wanted to fix the initial animation of the overview, we'd have to
always enable animations of the window clones when syncing geometries,
but that would break the animation of the workspace when hovering the
workspaceThumbnail slider, because right now those animations are "glued
together" using the actualGeometry, so they would get out of sync.
The reason there are no errors happening in workspace.js with the
existing code is that due to a bug in Clutter the fullGeometry of
WorkspacesDisplay gets set very early while mapping the WorkspacesViews
(because the overviews ControlsManager gets an allocation during the
resource scale calculation of a ClutterClone, see
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1181), so it
won't be set to null anymore when calling
WorkspacesView.animateToOverview().
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1119
The fullGeometry and the actualGeometry of the WorkspacesDisplay are set
from the allocation of the overviews ControlsManager and the
WorkspacesDisplay, that means they're only valid after those actors got
their allocations during Clutters allocation cycle.
Since WorkspacesDisplay._updateWorkspacesViews() is already called while
showing/mapping the WorkspacesDisplay, that allocation cycle didn't
happen yet and we end up either setting the geometries of the views to
null (in case of the fullGeometry) or to something wrong (a 0-sized
allocation in case of the actualGeometry).
So avoid setting invalid geometries on the views by initializing both
the fullGeometry and the actualGeometry to null, and then only updating
the geometries of the views after they're set to a correct value.
Note that this means we won't correctly animate the overview the first
time we open it since the animation depends on the geometries being set,
but is being started from show(), which means no allocations have
happened yet. In practice this introduces no regression though since
before this change we simply used incorrect geometries (see the 0-sized
allocation mentioned above) on the initial opening and the animation
didn't work either.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1119
clutter_text_get_selection() creates a copy of the selected text which
gets passed to st_clipboard_set_text() which creates its own copy. The
copy returned by clutter_text_get_selection() however never got free'd.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1306
(cherry picked from commit fdfcacf1db)
The shell mime sniffer goes through all the files in a directory,
however in case a file content type is not recognized, the GIO function
g_file_info_get_content_type() may return NULL, causing a crash when
looking up into the content type tables, as they are supposed to contain
strings only and they use `g_str_hash` has func, which doesn't support
NULL values.
So, in case we get an invalid content type, let's just ignore it,
without adding it to the cache as we do in the nautilus code that was
inspiring the sniffer.
Fixes https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/2643
(cherry picked from commit 46547ae027)
Spidermonkey caches imports, which means that uninstalling an
old extension version and installing a new one doesn't work as
expected: If the previous version was loaded, then its code will
be imported instead.
For the last couple of releases this has been a reliable source
of extension bug reports after major GNOME updates. Thankfully
chrome-gnome-shell removed its update support in favor of our
built-in support now, but users may still use older versions
or perform those actions manually, so it still makes sense to
catch this case and set an appropriate error.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1248
Unlike the desktop-entry hint, the app name is not optional. That
doesn't mean that we'll be able to match it to a .desktop file,
but we can at least try if we fail to match on PID or desktop-entry.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1291
(cherry picked from commit b487846c0a)
For fd.o notifications, we are taking the sender's PID into
account when associating notifications with sources (mainly
to deal with notify-send).
This broke when the implementation under the well-known name
was moved into a separate service, as the implementation in
gnome-shell will now always see the public notification-daemon
as sender.
Restore the old behavior by resolving the sender PID in the
separate service, and pass it as hint to the implementation
in gnome-shell.
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/2592
We currently only remove the screenshot operation from the shooter
map if the color pick operation completed successfully, but not if
it was cancelled. As a result, we now reject any further requests
from the same sender because we assume that there is an ongoing
operation.
Fix this by moving the cleanup to a finally clause that runs for
both code paths.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1288
Right now _createScreenshot() returns a tuple that indicates failure
when a sender requests a screenshot operation before a previously
started operation finished.
However that doesn't work for the PickColor() method, as it uses a
different return type than the other methods.
Address this by returning an error instead, which works in any case;
arguably trying to start multiple operations in parallel is an error
by the caller more than it is a failed operation anyway.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1288
As outlined in commit 36b8dcbe07, we can end up with wrong icons
if the icon theme changes right after a GTK theme change to/from
HighContrast triggered a theme reload.
That's because when we reload icons for the new icon theme, there
are already pending requests due to the icon-style change; those
requests are simply re-used for the new icons, with the existing
icon infos from the old theme.
The above commit applied a simple work-around by changing the
icon theme before the GTK theme, but that only works for the
HighContrast switch in our own UI.
It turns out that Settings also uses the "wrong" order, so the
issue still reproduces with the Universal Access panel.
So instead of relying on everything changing the settings in the
order we expect, cancel all ongoing requests on icon-theme changes.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1277
(cherry picked from commit d81237b9d6)
Since e06109c23c we keep old theme nodes
valid during the emission of the "custom-stylesheets-changed" signal.
It turns out that we might still look up the file of a stylesheet using
the files_by_stylesheet hashtable during the emission of that signal,
causing a crash because the assertion in _st_theme_resolve_url() fails.
So fix that and remove the stylesheet entry from the files_by_stylesheet
hashtable after emitting the "custom-stylesheets-changed" signal. And to
be consistent, also remove the entry from the stylesheets_by_file
hashtable after emitting the signal.
Fixes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/2757
(cherry picked from commit 0f947d4ff9)
Currently the switches handle position reflects the requested
extension state (as in: the user (de)activated the switch),
while the actual extension state is reflected by the underlying
state.
That doesn't work well when the switch is insensitive though (for
example on error), because the desaturation neuters the color
that reflects the state.
Address this by resetting the switch handle to the state when
making it insensitive.
https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/1263
(cherry picked from commit d2583aa47b)
