Jonas Dreßler [Fri, 3 May 2019 21:39:11 +0000 (23:39 +0200)]
layout: Show important actors above top_window_group
The top_window_group is used for windows like popup menus, which should
appear above shell chrome like the panel.
Since we want important actors such as the screen keyboard or modal
dialogs to be shown above those windows, add their actors after adding
global.top_window_group to this.uiGroup and provide a new function
addTopChrome() to add important chrome above the top_window_group.
st: Introspect ScrollViewFade to read fade properties
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.
As per the fact the menu item is an object we can just vfunc's.
This is not possible for events as specific union object is passed instead of
just the Clutter.Event.
popupMenuItem: Make it a widget inheriting from St.BoxLayout
Don't use composition for PopupBaseMenuItem, but instead inherit from BoxLayout.
So remove the internal actor, making all the menu items actor themselves.
Add an actor property as fallback to avoid warnings for usage in menus.
Florian Müllner [Sun, 4 Nov 2018 23:21:26 +0000 (00:21 +0100)]
systemActions: Only do prefix matches
Our search for system actions is currently inconsistent with searching
for applications: While we match terms anywhere within keywords, GIO
will only match at the beginning of words.
In order to get the same behavior, split keywords into single words
and only match terms at the beginning of a word.
js: Use GTypeFlags to define abstract GObject classes
gjs now supports an optional GTypeFlags value for GObject subclasses
defined with GObject.registerClass(), so it is not possible to define
abstract classes on the gobject-level, just like from C.
keyboard: Unset current page when it gets destroyed
The current page is going to be destroyed as part of _onKeyboardGroupsChanged
cleanup, however we don't unset its pointer, and thus we'd might try to call a
function using an invalidated pointer.
So, unset the this._current_page reference when its get destroyed, connecting
to destroy signal when setting it and disconnecting when changing page.
Florian Müllner [Thu, 2 May 2019 19:12:01 +0000 (21:12 +0200)]
st: Delay handling of :first/:last-child changes
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.
Fabrice Bellet [Tue, 7 May 2019 15:21:09 +0000 (15:21 +0000)]
network: Handle interface name changes
The interface name when a device is added may not be the final one. For
example when using USB tethering, it will first appear as 'usb0' before
being renamed to something like 'enp0s20f0u1' depending on the port the
phone is plugged in.
As a result, we will ignore the new interface name in that case and fail
to associate the correct connection with the device: Instead of the
correct "USB Ethernet" (or user-customized name), it will show up as
"Ethernet".
Fix this by updating names and connections when a device's interface
property changes.
Dylan McCall [Mon, 5 Feb 2018 18:46:22 +0000 (10:46 -0800)]
panel: Drag topmost maximized window
This makes it possible to drag a window which appears connected with the
panel, even if it is not in focus. As a result, it should be easier to
manipulate side-by-side windows.
texture-cache: Keep aspect ratio for content images
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
Jonas Dreßler [Fri, 3 May 2019 20:56:02 +0000 (22:56 +0200)]
windowManager: Fix fullscreen windows in ws switching animations
To prevent a small gap between windows in the workspace switching
animation, we temporarily shift windows up or down by the height of the
panel. This obviously breaks the animation for fullscreen windows, those
will overlap with the ones on the other workspace since there is no
panel shown in that case.
Fix this by checking whether the old or new workspace includes a
fullscreen window and don't shift the windows if there is one.
dialog: Really set ellipsize mode in subtitle and body
Dialog's subtitle or body could not be properly wrapped, while it's ellipsized
when the text's width doesn't exceed the container size.
Clutter text has an `ellipsize` property, however in dialog's subtitle and body
we have been setting the `ellipsize-mode` property to Pango.EllipsizeMode.NONE
that is not present in the underlying GObject.
Not being an error in javascript, gjs didn't warn us about this, while at the
same time the St.Label's default Pango.EllipsizeMode.END was used.
dashSpacer: Don't trigger allocations from size negotiations
If an actor's allocation is outdated, clutter_actor_get_allocation_box()
will queue a relayout. That's why it's advised to not use the function
unless the allocation is known to be valid (namely during paint), but
in particular not from within get_preferred_width/height vfuncs.
Using the :allocation property (which may be outdated) would be better,
but in this case we can simply delegate the request to the correct actor.
Carlos Garnacho [Tue, 30 Apr 2019 12:08:25 +0000 (14:08 +0200)]
keyboard: Destroy old layout actors when regenerating keyboard groups
We were cleaning up self._groups, but the actors for all previous
groups/layers/modes would remain attached to the aspect container,
simply hidden.
Under some circumstances this can really make the amount of actors
in the shell stage to quickly ramp up, it's not just a "leak" but
also has potential side effects on performance.
We should destroy all child actors of this._aspectContainer, except
the static ones (emoji and keypad).
While at it, fix this._groups re-initialization, as it's actually an
object, not an array.
worldClocks: Ignore locations with unknown timezone
We currently assume that every location has an associated timezone.
While this is sound in the real world, in practise it depends on
whether or not libgweather can find a corresponding timezone DB
entry.
This used to be a fringe case, but has become more likely when commit
https://gitlab.gnome.org/GNOME/libgweather/commit/d7682676ac9 moved
weather stations from cities to countries - the station itself is un-
likely to have a timezone entry, and the country may be part of more
than a single timezone.
It would be good for libgweather to return a timezone for those
locations again, but we should defend against the case anyway.
We cannot tell what time it is at a particular location without
knowing the timezone, so simply filter them out.
Carlos Garnacho [Tue, 30 Apr 2019 12:51:47 +0000 (14:51 +0200)]
st: Reload icon on style changes only if necessary
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.
Carlos Garnacho [Tue, 30 Apr 2019 12:42:59 +0000 (14:42 +0200)]
st: Use g_signal_handler_disconnect() to disconnect from StTheme
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.
The top bar handles allocating all its children itself, so there's
little value in chaining up to st_widget_allocate() and get the
default layout manager allocating all children again (and possibly
differently).
If this happens, we end up with an infinite allocation cycle with
corresponding performance penalty. Fix this by just doing and what
Shell.GenericContainer did before commit 286ffbe2b6 replaced it,
and not chain up to StWidget.
Carlos Garnacho [Fri, 26 Apr 2019 11:03:42 +0000 (13:03 +0200)]
padOsd: Use non-deprecated librsvg API to create handle
On one hand, we were using a path instead of an URI on
rsvg_handle_set_base_uri(). This broke at some point in librsvg
(presumably for the best, handling paths there sounds non-standard)
leaving a blank svg (As the base image wouldn't be accessed).
On the other hand, we use this with the deprecated rsvg_handle_write()
which we should drift away from.
Using rsvg_handle_new_from_stream_sync() neatly solves both. We use
newer API based on input streams and GFiles, and it internally does
the right thing, bringing the pad OSD back to life.
Unlike regular keys that generate key events from a virtual device,
emoji keys rely on the input method to insert the character. However
as the compositor cannot inject IM events into ibus, this only works
in the shell's own entries on X11.
We shouldn't expose mostly broken functionality to the user, so limit
the feature to the wayland session.
git-fetch's -q flag doesn't suppress warnings, so it's not a full
replacement of the redirection that was removed in commit 8cefd919.
Shut up the cryptic warning and replace it with a clearer log message
instead.
Depending on how gitlab's CI checks out gnome-shell, the shell branch
may not have a local reference like "gnome-3-32", but only a remote
one like "remotes/origin/gnome-3-32".
Consider that case as well when looking for a corresponding mutter branch.
If we don't find a branch that matches the branch used in the merge
request, we currently fall back to the non-merge-request matching,
i.e. first try the current shell branch, then fall back to master.
This should work for commits to upstream branches, but not for merge
requests to a stable branch. For those, the target branch name is
a better fallback.
Carlos Garnacho [Wed, 24 Apr 2019 10:24:12 +0000 (12:24 +0200)]
ci: Ensure we clone a deep enough history for commit review
It seems gitlab changed something recently in the default clone depth
which made MRs with >10 commits to obscurely fail in the review stage.
As per https://docs.gitlab.com/ee/ci/yaml/#shallow-cloning, bump it
to 100 to allow bigger MRs.
Simon McVittie [Tue, 16 Apr 2019 22:59:58 +0000 (23:59 +0100)]
ShellApp: Use g_signal_connect_object for window signals
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>.
Carlos Garnacho [Fri, 19 Apr 2019 13:43:39 +0000 (15:43 +0200)]
st: Only emit ::style-changed on actual changes
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.
environment: Warn when trying to access a deprecated actor property
The port away from the old Shell.GenericContainer actor turned many JS
classes with a corresponding actor property into Clutter.Actor subclasses.
For compatibility reasons, those properties were kept around for a while.
They were now removed and any code that still uses them should be adjusted.
Facilitate that transition by defining the compatibility property on
Clutter.Actor itself, but log a warning every time it is accessed.
Florian Müllner [Mon, 11 Feb 2019 18:55:42 +0000 (19:55 +0100)]
appMenu: Add heading to windows section
It may not be immediately obvious that the windows section is a
list of open application windows, as titles like "Downloads" can
easily be confused with an action. Add a section heading to avoid
confusion.
Florian Müllner [Mon, 11 Feb 2019 19:03:05 +0000 (20:03 +0100)]
appMenu: Remove separator in window section
We currently copy the app icon menu behavior, which puts a separator
between windows from the current workspace and windows from any other
workspace. It is more useful to have the windows section appear as a
clearly marked group, so drop the separator.
Carlos Garnacho [Sat, 23 Mar 2019 15:33:58 +0000 (16:33 +0100)]
st: Round CSS units to the nearest multiple of the scale factor
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.
verdre [Sat, 2 Feb 2019 16:01:29 +0000 (17:01 +0100)]
barLevel: Don't show border radius if the value is 0
Work around a known regression from [1] that caused the volume bar in
the OSD window to never be hidden, even if the volume is set to 0. This
happened because the border radius of the barLevel is always drawn
without ensuring that the actual bar is visible.
So simply check if the value to draw is 0, and if it is, don't draw the
border radius of the bar at all. This will still result in incorrect
representation of values that have a width smaller than 2*border-radius,
but at least the bar looks right for a width of 0 now.
When determining the biggest icon size that fits the available height,
we first subtract the additional space requirements of icons (spacing,
padding, running indicator etc.) and then divide the result by the
number of icons to get the maximum size available to each icon texture.
In the above, the additional space requirement of each icon is taken
from the first icon (as all icons are assumed to be the same), and
calculated as the difference between the icon button's preferred height
and the currently used icon size.
To make sure that the icon is actually using the dash's current icon
size (even while animating to a new icon size), we enforce its height
during the size request and restore its original height afterwards.
However after some recent changes, that step is causing troubles:
For some reason, the original height may be 0, and when we restore it,
we end up forcing a fixed non-height that bypasses the regular size
request machinery.
While it is unclear where exactly the zero height comes from (maybe
waiting for a valid resource scale?), it is clear that it's best
to avoid forcing a fixed height. So instead of making the icon
texture comply with the assumed icon size, adjust the calculations
to use its current height request.
dateMenu: Make clock offsets relative to local time
We recently added offsets to world clocks that represent the location's
timezone as UTC offset. However for most users, that representation is
overly technical and less helpful than the difference to their local time.
dateMenu: Make sure we always display a clock name
We currently use the city name for all location except named timezones.
However locations only have a city name if they are of level CITY or
DETACHED, or if they are of level WEATHER_STATION with a parent of level
CITY.
So when libgweather commit d7682676ac9 moved weather station locations from
cities to countries, it broke their names in the world clocks section.
To fix this, stop making assumptions about when we can use the city name
and simply try it first for all locations and fall back to the plain name
if its not available.
After the Adwaita refresh, the button shows up oval rather than
circular. To address this, make sure that the "image-button" class
is applied as well by using the dedicated setter function.
boxpointer: Compute source allocation and work area just once
Compute the source actor workarea and allocation when repositioning and keep it
cached so that we've not to calculating it again in _calculateArrowSide.
Since _calculateArrowSide only is called inside _updateFlip that is always
called just after reposition, we can be sure that the computed values are still
correct.
boxpointer: Use X/Y actor positioning instead of an anchor point
Currently all the widgets using BoxPointer and so popup actors are always drawn
at screen origin and then shifted at paint time in the proper position.
This doesn't work when using resource scale, since the widgets need to know
in which monitor they currently are in order to use properly scaled resources.
So, basically revert commit 22c22e0d7 and go back using the actual actor
coordinates for positioning and the actor translation for animating it, as the
relayouting issues of the past seem to be gone now.
boxpointer: Compute arrowStride using sourceActor monitor as reference
I suppose this was also the goal of commit 96e14dc, but because it was checking
the value of the always-undefined this.sourceActor, it really never worked as
expected.
So, now since we assume that this._sourceActor is always defined just use it
as reference to compute the monitor menu position.
windowManager: Don't move sticky windows when inserting workspace
Inserting a workspace is implemented by appending a new workspace, then
shifting all windows after the "new" workspace up. This has an unintended
side effect on sticky windows, as changing its workspace will unstick it.
Fix this by excluding sticky windows - there's little point in moving them
anyway, given that they should be on all workspaces (including the original
workspace and the target one).
Benjamin Berg [Thu, 11 Apr 2019 09:43:03 +0000 (11:43 +0200)]
shellDBus: Accept connector to restrict OSD to one monitor
The monitor index is not something that gnome-shell and
gnome-settings-daemon seem to be agreeing about. Using the connector
string is a much more reliable method of identifying a specific screen
and we are indeed using this already for monitor labling.
So switch over to use the connector rather than the monitor index. If a
user tries to use the old API, then the OSD will simply show up on all
monitors (which is the status quo currently anyway).
Benjamin Berg [Thu, 11 Apr 2019 17:37:15 +0000 (19:37 +0200)]
shellDBus: Remove ShowMonitorLabels API
This API has been broken for quite some time now as the corresponding
mutter function meta_monitor_manager_get_monitor_for_output was removed.
If anyone tries to use it, we would just run into a backtrace.
screenshot: Don't pick up rubberband style from GTK
Selecting a screen area for a screenshot isn't the same as selecting
items in an icon view, so there's no strong rationale for picking
up the style from GTK. We stopped doing that for other elements like
tile previews long ago, so just use our own style here too.
Florian Müllner [Wed, 20 Mar 2019 17:19:01 +0000 (17:19 +0000)]
workspacesView: Support touch/touchpad gestures to switch workspaces
Now that the existing touch/touchpad gestures in windowManager only
handle normal mode, add corresponding gestures for the overview and
hook them up to the existing workspace scroll animations.
Florian Müllner [Wed, 20 Mar 2019 15:21:01 +0000 (15:21 +0000)]
windowManager: Limit switch gestures to normal mode
The window group is hidden while in overview, so the stick-to-content
animation isn't visible either. Worse, the gestures messes up the
position of window actors in that case. Just limit the gesture to
normal mode for now, we will soon add it back in the overview with
its own animation handling.
projects / gnome-shell.git / log