Delay login animation until after wallpaper has loaded
Currently, the login animation can occur before the user's wallpaper has
been loaded. When this happens, we wind up displaying a solid blue
background for half a second or so before the proper background is
displayed. This looks jarring and bad. It's great that we can start
GNOME quickly, but starting up before the wallpaper is ready is *too*
quickly.
I've been meaning to fix this since 2014. Better late than never! We can
just have BackgroundManager emit a loaded signal the first time it loads
its first background, and have the startup animation code wait for that
before proceeding.
Some of this code is by Florian, who helped with promisifying. Thanks!
On Wayland, navigating menus with the keyboard would not open drop-down
menus when NumLock is enabled.
That's old issue (gnome-shell#550) that was not completely fixed with
commit 88556226 because the lock mask needs to be filtered out in
_onKeyPress() as well.
appDisplay: Append new icons at the first available page after first
As per design discussion, the first page is a somewhat of a special
page where we really don't want to change anything unless necessary.
Append new icons at the first available slot after the first page.
Make the placeholder icon be appended to the first available page
as well, since it's always used when dragging from folder dialogs.
appDisplay: Use a drag monitor to check for out-of-dialog drags
When the app folder dialog handles a drag hover, it starts a timeout
to popdown if dragging outside the "real" dialog area. However, when
dragging inside it, BaseAppView handles all drag hover events which
would disarm the popdown timeout. In cases like this, it's almost
impossible to prevent the timeout from triggering, which always pops
down the dialog.
Add a drag monitor when handling any drag hover (which only happens
when dragging outside the folder's icon grid); and eventually disarm
the popdown timeout from the monitor's motion event. Remove the drag
monitor when dragging over the folder dialog again.
Now that the DnD code is shared between AppDisplay and
FolderView, we hit an unexpected problem: FolderView is
handling drag events even when the folder dialog is hidden.
As a side effect, this spams the journal with warnings.
Only handle drag events when mapped. On unmap, disable
the view's drag monitor, and disconnect from all drag
events.
appDisplay: Ensure icons always are at the correct position
When redisplaying, we currently only remove and add icons, but
never adjust the position of already added icons. If the icon
position changed, it wouldn't be reflected on the icon grid.
Now that the icon being dragged can come from AppDisplay and also
a folder dialog, the check for when to overshoot is broken. Check
if the icon is a BaseAppView icon.
appDisplay: Use custom function to retrieve item page and position
It is important that '_loadApps()' return a sorted list -- adding the
same icons at the same positions but in different orders results in
a wrong icon grid.
Add support for using a custom positioning function, and implement it
in AppDisplay. Because FolderView doesn't implement a custom sorting
function, the items are still sorted alphabetically.
This is the key that will be used to store the pages and the icons in
each page. The idea is that we we store an a{sv} variant for each page.
This variant will contain <icon id> → array of properties, where we
can store arbitrary data for each icon. The expected output of this
key is:
The leeways are parts of the icon that ignore incoming drag
events. This is how IconGrid and IconGridLayout treat it, and
this is how the icons should treat themselves too.
Make AppIcon ignore dragging over the left and right leeways.
Sebastian Keller [Mon, 20 Jul 2020 10:05:15 +0000 (12:05 +0200)]
calendar-server: Remove delay before event emission
The timeout seems to have been carried over from the old code that
relied on gnome-shell calling 'GetEvents' after every 'Changed' signal
where it was used to throttle the signal. In the new code where
calendar-server is sending the changes themselves via signals this is no
longer necessary and actually causes a delay when switching between
months.
While GtkIconTheme does look up icons in the toplevel icons resource
path, it will only use them as ultimate fallback. That is, if the
icon theme (or the hicolor fallback) include a "keyboard" icon, it
will be used over the "keyboard-enter-symbolic" icon in the resource.
Moving the icons to appropriate subdirectories gives them higher
priority than the fallback names, and thus fixes the issue.
extensionDownloader: Fix check for updates with several extensions
When having several extensions installed checking for updates fails.
This is because we are using GET and query params and since we are
sending all the metadata of the extension the server returns 502
when the URL is too long. This error code is ignored safely.
It is only needed to send the version of the extension to check if it
has updates.
Jonas Ådahl [Thu, 9 Jul 2020 07:18:27 +0000 (09:18 +0200)]
layout: Remove redundant background refresh
When using the NVIDIA driver, textures tend to loose their pixels when
suspending. In the past we handled this by figuring out when the NVIDIA
driver was used, and reload the background whenever we noticed we
resumed from suspend.
This shouldn't be needed anymore after
https://gitlab.gnome.org/GNOME/mutter/merge_requests/600, as it should
handle this by listening to video-memory-purged signal. Thus remove our
special handling here.
workspace: Always set state at the end of overview transition
We don't change to the zoomed out state when fading to the overview,
however we should do that after the transition in case the user
switches to the window picker.
workspace: Reset window opacity after overview transitions
When using the fade animation when transitioning to the overview instead
of zoom, we fade out all window previews to fully transparent. But after
commit 751189253a removed the old _updateWindowPositions() function,
nothing resets the opacity again, so when switching from the app- to the
window picker, all previews are hidden.
Fix this by always resetting the window preview opacity after showing
or hiding the overview.
workspace: Always use floating layout for fade transition
We don't animate size and position when fading, so we want all previews
to already be at their final position. However when the app picker is
opened from within the overview, window previews use the zoomed layout,
so that's the state we are then fading when leaving the overview from
the app picker.
Fix that by setting the correct state at the start of the fade transition.
(In the case of fadeToOverview(), the value should always be correct
already, but set it anyway for symmetry with fadeFromOverview())
workspacesView: Really don't animate primary view when fading
In commit 9297d87775d64 we stopped syncing the primary view's actual
geometry at the start of the transition when doing a fade animation,
however the view animation may still be triggered by an allocation
change.
Prevent those unwanted size changes during fade by keeping track of
the fade state and explicitly skip syncing the geometry while a fade
is ongoing.
workspacesView: Do not skip entire overview transition
Since commit af543daf1c3, we skip the overview transition when the
actual geometry hasn't been set yet. However with the new layout
manager, the only bit that still needs the separate geometry is
the transition of the view, the workspaces can do their transition
just fine.
workspacesView: Update visibily when gesture drag begins
When dragging the workspaces through the swipe gesture, all
workspaces must be visible. WorkspacesView's _updateVisibility()
method special-cases this and ensures that.
However, this method is only called when (1) going to the active
workspace, and (2) when the gesture ends. That means, if there
is any workspace hidden by the time a gesture starts, it is never
shown!
Call _updateVisibility() on startTouchGesture() as well.
shell/window-tracker: Tighten sandbox ID prefix check
Since commit b60836932 we only allow WM_CLASS matches for sandboxed
applications if the found app's ID is prefixed by the sandbox ID.
The existing check still has a hole in it though: "org.example.Foo"
and "org.example.FooDevel" are different applications, yet the former
is a prefix of the latter.
So tighten the check by including a trailing "." in the checked prefix;
this excludes cases like the above, while still working for the regular
case of a single .desktop file because our app IDs include the ".desktop"
suffix.
Jonas Dreßler [Fri, 19 Jun 2020 12:13:12 +0000 (14:13 +0200)]
windowPreview: Adapt to ClutterActor API changes
The API of clutter_actor_allocate_preferred_size() changed with
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1310, so adapt to
this change and pass the origin to the function.
workspacesView: Do not animate primary view geometry when fading
When going straight to the app picker, we fade in the overview instead
of doing the full-blown zoom transition. In order to keep windows at
their floating position, we must apply the same to the view itself
and not transition to the overview geometry when fading.
workspacesView: Decouple syncing geometry from updating views
This gives us more control over when the geometry is synced. We
will soon use that to keep the primary view at the work area
geometry while fading to the overview.
Florian Müllner [Thu, 25 Jun 2020 17:54:54 +0000 (19:54 +0200)]
workspacesView: Add back overview transition
The transition was temporarily removed when switching to the new
workspace layout manager. Now everything is in place to reimplement
it with a combination of the layout manager's state adjustment and
the view's allocation.
Florian Müllner [Fri, 26 Jun 2020 14:04:44 +0000 (16:04 +0200)]
workspacesView: Properly allocate workspaces
So far we've been allocating workspaces in a stack, and relied on
translation to move them to the right position. And as the position
depends on both the workspace's index and the view's viewport, some
care is needed to prevent gestures/scrolling from interfering with
layout updates.
Clean that up by properly allocating workspaces in a row or column,
and use a translation to reflect the current scroll position.
Florian Müllner [Thu, 25 Jun 2020 23:08:29 +0000 (01:08 +0200)]
workspacesView: Use work area for default geometry
The workspace's layout manager keeps the workspace at the same ratio as
the work area, so it makes more sense to base the views' default geometry
on that as well than the monitor area we are using right now.
(It shouldn't matter much in practice, as this only affects views on
non-primary monitors where the work area usually matches the monitor area)
Florian Müllner [Fri, 26 Jun 2020 00:56:35 +0000 (02:56 +0200)]
workspace: Account for window picker padding
Workspaces are supposed to request the work area as their preferred
size, however the widget will adjust the sizes returned by the
layout manager to account for borders and padding.
Florian Müllner [Fri, 26 Jun 2020 15:01:31 +0000 (17:01 +0200)]
windowPreview: Allow to disable overlay
WindowPreviews now contain and manage overlaid elements like close
button or title label themselves. That's generally better, but right now
the only way to disable those overlays (for example during transitions)
is to prevent any hover or focus events from getting to the preview.
Instead, add some explicit API for enabling or disabling overlay support.
Florian Müllner [Thu, 25 Jun 2020 18:25:22 +0000 (20:25 +0200)]
workspace: Handle minimized windows in layout
The new layout manager always uses the window's regular size and position
for the floating state, but the preview of minimized windows should be
hidden instead.
Jonas Dreßler [Mon, 6 Jul 2020 13:44:25 +0000 (15:44 +0200)]
windowPreview: Ensure style as soon as preview is realized
For the windowPreview we need to ensure the style information of the
border and title is up-to-date when chromeWidths() or chromeHeights() is
called. Since the introduction of the WorkspaceLayout those functions
may be called during an allocation cycle, which means we should avoid
calling queuing relayouts inside them. Calling StWidgets ensure_style()
method will queue a relayout though in case the newly generated theme
node has a different geometry.
So avoid queueing a relayout during allocation cycles (and the warning
Clutter logs because of that) by ensuring the style of the border and
title earlier, as soon as the WindowPreview is attached to a stage.
Jonas Dreßler [Wed, 3 Jun 2020 16:17:54 +0000 (18:17 +0200)]
workspace: Use the new WorkspaceLayout for allocating window clones
Switch to the new WorkspaceLayout layout manager to allocate the window
clones of the overview properly using Clutters layouting mechanisms.
Since we now no longer make use of the fullGeometry, we can remove the
setFullGeometry() function from the Workspace class. Also we can stop
setting the actualGeometry on the Workspaces and WorkspaceViews and
instead just set the fixed position and size of the views to their
full or actual geometry. This also has the benefit that we no longer
have to set a custom clip, but can simply enable clip_to_allocation.
The geometry needs to be set inside a BEFORE_REDRAW later because
_updateWorkspacesActualGeometry() is called from a notify::allocation
handler.
This isn't doing any animations when showing/hiding the overview yet,
we'll add that in the next commit.
Jonas Dreßler [Wed, 3 Jun 2020 16:10:09 +0000 (18:10 +0200)]
workspace: Introduce layout manager for allocating the Workspace
Add a new ClutterLayoutManager for layouting the workspaces of the
overview, WorkspaceLayout.
This layout manager integrates the existing LayoutStrategies used to
layout the window clones of the overview and supports freezing the
layout, animating between layout changes and adjusting the spacing for
the width and height of the window chrome. It also adds support for a
layout of the windows that looks the same as the actual workspace,
transitioning between that layout and the LayoutStrategy can be done by
setting the value of an StAdjustment, available using the
stateAdjustment getter function.
This will replace the current static-positioning based layouting of the
window clones in the next commit.
Jonas Dreßler [Wed, 3 Jun 2020 16:07:01 +0000 (18:07 +0200)]
workspacesViews: Use translation for showing Workspaces
We're going to use fixed position for positioning workspaces when
they're allocated by their own layout manager, using those positions to
scroll between different workspaces interferes with that, so do that
using translations instead.
keyboard: Request a bigger size in portrait orientation
In portrait orientation, we set the height to the preferred height
for the monitor width (or, if smaller, a third o the screen height).
However as the forWidth currently doesn't make a difference, the height
is effectively controlled by the natural height of the keys - which is
rather small.
Address this by making AspectContainer request an appropriate preferred
size based on the fixed ratio.
Currently our heuristics for matching a window to its app check
for the application ID before the WM_CLASS, as the ID is more
reliable in so far that it is outside the application's control
and so it cannot use it to spoof a different application.
However this also prevents applications with multiple .desktop
files like LibreOffice from matching any .desktop files other
than the one under the main ID.
Since we now no longer allow the WM_CLASS to match a .desktop
file that doesn't belong to the sandboxed application, we can
fix that issue by checking the WM_CLASS first, without opening
the door to spoofing.
gdm: Introduce vmware credential manager for pre-authenticated logins
The previous commit implemented a new CredentialManager interface to
facilitate adding additional providers for pre-authenticating the user
at the login screen.
This commit implements a new credential manager using that interface
for vmware deployments.
yun341 [Fri, 3 Jan 2020 16:31:15 +0000 (00:31 +0800)]
gdm: Refactor oVirt to a generic CredentialManager interface
Commit 4cda61a1 added support for pre-authenticated logins in
oVirt environments. This feature prevents a user from having
to type their password twice (once to the oVirt management machine,
and then immediately again in the provisioned guest running gnome-shell).
That feature is currently oVirt specific, but a similar feature would
be useful in non-oVirt based virt farm environments.
Toward that end, this commit generalizes the various aspects of the
oVirt integration code, so that it can be reused in a subsequent
commit for adding single sign on support in vmware deployments, too.
Florian Müllner [Tue, 30 Jun 2020 14:53:50 +0000 (16:53 +0200)]
shell/app: Add new get_icon() method
Now that we can always associate a GIcon with the app, add a method
to access it. While create_icon_texture() is still likely to be more
convenient in most cases, exposing the icon can still be useful, for
example to add it to a different kind of actor or to compare it with
other GIcons.
Florian Müllner [Tue, 30 Jun 2020 14:48:16 +0000 (16:48 +0200)]
st/texture-cache: Return a GIcon from bind_cairo_surface_property()
We still load the surface into an StImageContent, but instead of
adding the content to an actor we hand out, return the content
itself (as GIcon).
That means we lose the ability to specify an icon size, but as we
get the pixel data from a fixed-size surface anyway, that shouldn't
matter much in practice.
Not to mention that the function is only used for fallback X11 icons,
which are already shit more often than not.
Florian Müllner [Tue, 30 Jun 2020 02:12:16 +0000 (04:12 +0200)]
st/image-content: Implement GIcon interface
On the one hand, this is a bit of a stretch: StImageContent is what
we create from GIcons.
But on the other hand, there's some justification: StImageContent does
represent an image (and likely icon) after all, and there's some
precedent with GdkPixbuf.
In the end as we don't care about serialization or loading from other
API, we can go with a very crude implementation that allows us to
pass out a content as GIcon and use it directly when "loading" it.
We will use that soon to represent X11 window icons as GIcons, which
in turn will allow us to unify app icon handling.
Jonas Dreßler [Thu, 12 Mar 2020 12:33:25 +0000 (13:33 +0100)]
st/viewport: Invalidate transform when it changes
Since ClutterActor now caches the whole transformation matrix for an
actor, we need to invalidate the cached transform if the matrix returned
by apply_transform() implementations changes.
Jonas Dreßler [Thu, 28 May 2020 12:47:17 +0000 (14:47 +0200)]
st/widget: Remove get_resource_scale function
ClutterActor provides the same function, but with a different return
value. So since we already switched to the ClutterActor implementation
in our C code, we can now safely remove st_widget_get_resource_scale()
and update the JS code that's still using the old API.
Jonas Dreßler [Thu, 28 May 2020 12:43:26 +0000 (14:43 +0200)]
Switch to ClutterActors resource-scale-changed signal
Instead of using the "notify::resource-scale" signal and StWidgets
"resource-scale-changed" signal, use the new "resource-scale-changed"
signal of ClutterActor, which replaces its "resource-scale" property.
Since we'd now have two "resource-scale-changed" signals, one on
ClutterActor and one on StWidget, remove the StWidget one in favour of
the new one.
Florian Müllner [Tue, 30 Jun 2020 11:08:03 +0000 (13:08 +0200)]
unlockDialog: Fix scale-factor handling on multihead
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.
Sebastian Keller [Mon, 29 Jun 2020 20:51:41 +0000 (22:51 +0200)]
altTab: Remove down arrow when removing an app from switcher
The arrow of the removed app was still left in the list with the
visibility of the arrow still depending on the original list order. This
could either lead to apps with just one window now suddenly having a
down arrow or apps with multiple windows not having one. If the last
window in the list had a down arrow, it would have been displayed
outside the window switcher.
projects / gnome-shell.git / log