extensionSystem: Add method for opening extension prefs
Extension that want to expose their own preferences (for example as menu
items) do that by passing their UUID to gnome-shell-extension-prefs.
But since 3.36.1 the app is optional and no longer accepts arguments on
the command line. To adjust, extensions now need to make a D-Bus call
the extensions portal, just like the app and gnome-shell.
We will add a convenience method for that purpose, so it makes
sense to share the existing code. As it's extension-related, the
extension manager looks like the right place ...
Florian Müllner [Thu, 19 Dec 2019 19:50:37 +0000 (20:50 +0100)]
js: Account for promisified call() method
A promisified method expects the callback parameter to be either
a function (in which case the original method is called normally)
or omitted altogether (in which case a Promise is returned).
The call to open application details in Software does neither and
passes null instead, which will result in a warning (because no
function argument means a promise will be used, but not omitting
the parameter means we end up with too many arguments).
The app-cache code currently stores the folder translations in a hash
that can be accessed via shell_util_get_translated_folder_name().
This hash uses the filename (inc. extension) for the "desktop-directory"
as key which causes an issue when trying to find the translation
on AppDisplay._findBestFolderName() which gets categories (folder names)
from the app info which doesn't contain the ".directory" extension.
Fix that by storing the filename without extension as the hash key for
the cached folder translations.
Mariana Picolo [Mon, 30 Mar 2020 08:48:06 +0000 (05:48 -0300)]
bluetooth: Add getDeviceInfos function
In case where only one device is connected, we want to display its name
in the menu. For that we will need more than the number of known/connected
devices, so change the function to return an array of device infos instead.
main: Don't override DesktopAppInfo desktop if already GNOME
During the shell initialization we call the (deprecated) function to
override the Desktop environment in Gio DesktopAppInfo to make sure that
applications are correctly shown (as per commit b2fbf5a2), however this
might break the cases in which $XDG_CURRENT_DESKTOP is already set and
contains GNOME (given that is now a list).
In Ubuntu this is in fact set to: ubuntu:GNOME.
Now, if an application contains NotShowIn=ubuntu, the key will be ignored by
the shell, and the application is still listed everywhere.
So, override the DesktopAppInfo desktop environment only in the case that
the current desktop is not already GNOME.
At the current date I think we could just safely get rid of this override at
all, but there could be still cases where it still might be useful, like when
running as nested in some other environment, so keeping it.
Jonas Ådahl [Tue, 31 Mar 2020 17:00:58 +0000 (19:00 +0200)]
appDisplay: Clear animateLater callbacks when unmapping
In some situations we could end up not with lingering 'view-loaded'
handler. This could result in delayed spring animate-in being initiated,
e.g. after a minute after the activities overview was already closed.
Fix this by removing any lingering signal or later handlers when
unmapping.
Florian Müllner [Thu, 19 Dec 2019 19:50:37 +0000 (20:50 +0100)]
js: Promisify async operations
Promises make asynchronous operations easier to manage, in particular
when used through the async/await syntax that allows for asynchronous
code to closely resemble synchronous one.
gjs has included a Gio._promisify() helper for a while now, which
monkey-patches methods that follow GIO's async pattern to return a
Promise when called without a callback argument.
Use that to get rid of all those GAsyncReadyCallbacks!
Jonas Ådahl [Mon, 30 Mar 2020 13:16:53 +0000 (15:16 +0200)]
appDisplay: Don't start animation from the 'paint' signal
Starting the animation from the actor 'paint' signal has various
unwanted consequences, such as sometimes trigger a
clutter_actor_queue_relayout() during the paint phase. One unwanted
consequence was that an offscreen actor effect was disabled during
painting, meaning the effect would begin being active, but later during
the post-paint processing being disabled. The caused said effect to push
an offscreen framebuffer to the paint context, but then just destroy it
instead of popping it. When this happened, we'd end up trying to operate
on a framebuffer that may had been finalized, or not, depending on the
garbage collector. Sometimes, for some users, this caused a segmentation
fault when trying to pop a matrix from the framebuffer matrix stack.
Deal with this more properly, by using the 'view-loaded' signal to wait
with animation until the view is loaded, as well as using MetaLater to
schedule the start of the animation.
For when a view was signalled to be ready, we're in a state where we can
start animation before the next frame as the layout is ready, but when
not, we have to add back the "hack" where we must wait for one frame for
the target icon positions to be up to date. Do this by adding a
MetaLater IDLE callback that starts the animation *after* the next
frame. This also needs the old 'opacity = 0' work around to not show an
incorrect first frame.
Florian Müllner [Wed, 18 Mar 2020 21:20:33 +0000 (22:20 +0100)]
js: Do not set entries' ClutterText:editable property
The property influences the text's (and thus entry's) minimum width[0],
which is generally not what we want. And as we now prevent text from
being entered in non-reactive entries by other means, we can simply
drop it.
Florian Müllner [Sun, 29 Mar 2020 17:36:12 +0000 (19:36 +0200)]
st/entry: Unset key focus when made unreactive
It seems reasonable that an entry shouldn't allow entering text when not
reactive. The same could be achieved by changing the text's :editable
property, however that will disable scrolling if the text doesn't fit,
which may result in an unwanted size change.
Florian Müllner [Wed, 18 Mar 2020 21:10:36 +0000 (22:10 +0100)]
st/entry: Bind ClutterText reactivity to entry
The text is part of the entry, so it is surprising that it can
still be edited when the entry itself isn't reactive. Address
this by setting up a binding instead of expecting all consumers
to handle the case themselves.
Luke Yelavich [Sat, 28 Mar 2020 03:56:23 +0000 (14:56 +1100)]
calendar: Improve the accessibility of the "Do Not Disturb" switch
Set the do not disturb label as the label actor for the do not disturb switch,
so that Orca speaks the do not disturb label when the user moves
keyboard focus to the do not disturb switch.
Also enable toggle mode for the "Do Not Disturb" button and bind it's checked
state to the state property of the switch. This makes sure that Orca presents
thecorrect state of the do not disturb switch to the user.
Jonas Dreßler [Sat, 28 Mar 2020 13:36:20 +0000 (14:36 +0100)]
keyboard: Hide keyboardBox after destroying the keyboard
It seems there is a weird race condition between Clutter trying to
destroy the keyboard actor and Clutter trying to hide the keyboardBox
container actor: If the keyboardBox is hidden before destroying the
keyboard actor, Clutter doesn't repaint anything and the keyboard
remains visible until something else draws over it.
To fix this issue until we find the underlying Clutter bug, simply
destroy the keyboard actor before hiding the keyboardBox. The order in
which we call these doesn't matter anyway since hideKeyboard(true) hides
the keyboard immediately without an animation.
Jonas Dreßler [Sat, 28 Mar 2020 13:34:24 +0000 (14:34 +0100)]
layout: Use translation_y of 0 to hide keyboard
Since we show the keyboard using a translation_y of -keyboardHeight, the
keyboard will be moved down far enough to be out of sight by setting
translation_y to 0.
Jonas Dreßler [Sat, 28 Mar 2020 13:15:09 +0000 (14:15 +0100)]
keyboard: Don't include keyboard devices when updating lastDevice
We're dealing with attached keyboards now using the touch_mode property
of ClutterSeat: If a device has a keyboard attached, the touch-mode is
FALSE and we won't automatically show the OSK on touches, also the
touch-mode gets set to FALSE when an external keyboard is being plugged
in, so that also hides the OSK automatically.
With that, we can now ignore keyboard devices when updating the last
used device and no longer have to special-case our own virtual devices.
Because there was no special-case for the virtual device we use on
Wayland now, this fixes a bug where the keyboard disappeared after
touching keys like Enter or Backspace.
Florian Müllner [Sat, 28 Mar 2020 02:37:50 +0000 (03:37 +0100)]
volume: Cancel before checking state
Since commit 2894085c45 we omit sound feedback on volume changes
if something is already outputting sound. Unfortunately that
"something" may be our own feedback (from a previous volume
change).
In that case we do not want to omit the new feedback, so instead
cancel the previous one.
Jonas Dreßler [Sat, 28 Mar 2020 10:54:32 +0000 (11:54 +0100)]
keyboard: Run dispose manually on virtual input device when destroying
We want to make sure any buttons that are still pressed on the virtual
input device used by the OSK are released immediately when destroying
the OSK. Do this by calling run_dispose() on the destroy() function of
the KeyboardController, which makes sure we don't have to wait for the
garbage collection to dispose the object and a still pressed key remains
being pressed until the GC kicks in.
Carlos Garnacho [Sat, 28 Mar 2020 15:38:14 +0000 (16:38 +0100)]
st: Honor alpha range specified by pango_attr_foreground_alpha_new()
This is documented as a value between 1 and 65536. However we were passing
a 0 value for 100% transparent colors, which is interpreted as "system
inherited" in pango_renderer_get_alpha() docs.
Ensure we respect this range by specifying the minimum allowed alpha (1)
if the color is fully transparent. If someone notices this 1/65535th change
I'll ask him how many pleiades can he count.
Carlos Garnacho [Sat, 28 Mar 2020 16:26:39 +0000 (17:26 +0100)]
inputMethod: Protect for running with older mutter versions
The offset argument is changing from uint to int. Which means we
might would pass a negative offset and trigger an "out of bounds"
error. Make it work more or less alright with older mutters, by
clamping the offset to 0.
AuthPrompt is created on demand, and this._authPrompt is
expected to be null except on very strictly controlled
occasions. The idle monitor callback isn't one of them.
Florian Müllner [Sun, 1 Dec 2019 02:28:56 +0000 (03:28 +0100)]
extensions-app: Add flatpak manifest
We finally have everything in place for distributing the Extensions app
as flatpak without jumping through too many hoops. Add a manifest that
can produce a nightly build like other GNOME modules, and can serve as
a template for a stable manifest on flathub.
Florian Müllner [Sun, 1 Dec 2019 02:01:02 +0000 (03:01 +0100)]
extensions-app: Move Extensions app to new subproject
The Extensions app code is now independent enough from the rest of the code
base to move it to its own subprojects, like we did for the extensions-tool.
This allows for stand-alone builds of the app, which we are about to use
for distributing it as flatpak.
Florian Müllner [Mon, 9 Mar 2020 00:59:09 +0000 (01:59 +0100)]
extensionPrefs: Fake Config module
We include config.js because it is a dependency of ExtensionUtils,
but it's not actually used in the code paths we exercise.
As we want to allow stand-alone builds of the app, it is much easier to
fake the module than to either include a generated file from elsewhere
in the tree or generate it ourselves.
Jonas Dreßler [Tue, 29 Oct 2019 23:06:10 +0000 (00:06 +0100)]
appDisplay: Only use dragMonitor for one icon at a time
Instead of adding a dragMonitor for every icon in the grid as soon as
one icon is getting dragged, only add a dragMonitor for the icon that is
currently being dragged over (ie. the current drag-target). With a large
number of icons in the iconGrid, this should significantly reduce lags
while dragging.
We can do this by detecting the DnD-entering of an icon or folder using
the `handleDragOver()` callback of drag-targets, adding the dragMonitor
because we know an icon is hovering above the drag-target and then
detecting the DnD-leaving of the drag-target by using the `dragMotion()`
handler, where we remove the dragMonitor again as soon as the
targetActor is no longer our actor (ie. the drag-target).
Jonas Dreßler [Sat, 23 Nov 2019 18:57:02 +0000 (19:57 +0100)]
appDisplay: Remove dragMonitor when FolderIcon is destroyed
While it should be impossible to destroy a FolderIcon while a DnD action
is still going on, there might still be rare cases where this happens
(ie. when a folder is removed because an app got deleted during DnD).
So make sure we're on the safe side here and don't potentially leave
dragMonitors around after the icon is destroyed by removing the
dragMonitor inside the onDestroy handler of the FolderIcon, just like we
do for AppIcons.
Jonas Dreßler [Thu, 13 Feb 2020 21:50:40 +0000 (22:50 +0100)]
blur-effect: Handle failure of background blitting gracefully
`paint_background` already provides a return value in case the blitting
of the framebuffer fails, handle that and fall back to only drawing the
actor in case something goes wrong.
We want to completely clear the background framebuffer when switching
back to ACTOR mode to make sure the `background_fb.framebuffer` check
will fail in `update_background_fbo` when switching to BACKGROUND mode
again. Otherwise the checks in `update_background_fbo` will return TRUE
and we will keep using the background framebuffer that was created
before switchig to ACTOR mode.
While at it, also clear the background framebuffer completely when
changing the actor to avoid the same issue here.
Jonas Dreßler [Wed, 12 Feb 2020 22:02:14 +0000 (23:02 +0100)]
blur-effect: Fix framebuffer sizes when stage-view scaling is used
When blurring only the actor (ACTOR mode), we don't want to apply any
scale, it looks fine without using the resource scale and it also seems
like `clutter_actor_continue_paint` in `paint_actor_offscreen` only
draws an unscaled texture anyway (ie. if the resource scale is 2, only a
quarter of the framebuffer is being drawn to).
In BACKGROUND mode though, we need to scale the framebuffer using the
scale factor of the stage view (ie. the final scale factor for the
monitor) because the content of the framebuffer we blit is scaled using
that factor. Also, since the framebuffer we blit belongs to a stage view
and only includes the contents of this view, we need to adjust the
stage-coordinates of the actor to be relative to the stage-view.
To make sure we don't have to get the transformed actor size or position
multiple times during one paint-run and don't have to carefully floor()
or ceil() widths and positions, store the size of the actor (which is
also the size of the framebuffer) and its position relative to the stage
view inside a ClutterActorBox.
Jonas Dreßler [Fri, 27 Mar 2020 22:00:36 +0000 (23:00 +0100)]
blur-effect: Don't check whether effect is enabled when painting
A ClutterEffect is being painted as part of the paint cycle of
ClutterActor, where _clutter_effect_paint() is called before painting
the actual actor. With that, it's impossible that an effect gets painted
while it's disabled, so remove the check whether the ClutterActorMeta is
enabled before painting.
Also if everything works fine in Clutter, the ClutterActorMeta should
have an actor set and we've been notified about that actor in
shell_blur_effect_set_actor(), so assert that our cached actor is set
when painting the effect.
Takao Fujiwara [Wed, 14 Aug 2019 08:16:46 +0000 (17:16 +0900)]
inputMethod: Fix delete-surrounding-text signal
Forward the arguments at the 'delete-surrounding-text' signal
from IBusInputContext to clutter_input_method_delete_surrounding()
so that ibus-typing-booster use the deleting surrounding text function.
Input method engines can delete the output text in applications
with this function.
This change will require a change of mutter of mutter!517
because the first arguemnt of the 'delete-surrounding-text' is INT
to express the offset of the current cursor position but
the first one of clutter_input_method_delete_surrounding() is UINT
since the Wayland spec accepts UINT in delete_surrounding()
mutter will change the type of the first one to INT in
clutter_input_method_delete_surrounding() to work with this change.
Florian Müllner [Fri, 27 Mar 2020 15:33:48 +0000 (16:33 +0100)]
cldr2json: Do not overwrite existing files
We had various requests to improve existing OSK layouts, but
haven't accepted them so far as any changes would be overridden
when regenerating the layouts.
However as the upstream layouts at http://www.unicode.org are
extremely slow to update(*), we shouldn't block all improvements.
So instead of letting the update script override all existing
layouts, just make it import new layouts.
(*) not their fault, as the android layouts are a downstream to Google
Florian Müllner [Fri, 27 Mar 2020 15:36:04 +0000 (16:36 +0100)]
data: Don't clone cldr2json when updating OSK layouts
Meh, we imported the module so we don't have to go through another
upstream to make changes to the script, but then ended up not using
the fork at all.
Florian Müllner [Fri, 27 Mar 2020 15:35:00 +0000 (16:35 +0100)]
data: Move cldr2json fork into subdirectory
The module was imported into the toplevel in !424, but that's at
least a bit weird:
- it's a helper script for one particular aspect (OSK layouts)
- it adds a README.mdwn to our own README.md, and a test/ directory
to our tests/
Move the whole thing to a subdirectory under data/, which is more
appropriate.
Jonas Dreßler [Thu, 26 Mar 2020 21:12:39 +0000 (22:12 +0100)]
theme/switcher-popup: Increase contrast of switcher items
Use a brighter color and increase the contrast of the selected/active
items in the switcher popup.
Since a bright color doesn't go well with a box-shadow, remove that
shadow, an effect like should only be used for elements clicked with a
mouse anyway.
Florian Müllner [Wed, 11 Mar 2020 19:09:11 +0000 (20:09 +0100)]
extensionSystem: Require Extensions app for updates
While we don't technically need the app to download and apply
updates, we do require it for notifying the user about available
updates and listing extensions with pending updates.
So instead of intransparently applying updates in the background
without the user noticing, disable updates altogether if the
Extensions app is not installed.
Blacklist support was added all the way back in commit 1e286e43, but
the code had been defunctional until recently. While uninstalling an
extension that has been blacklisted makes sense off-hand, unfortunately
we don't know if an extension was *actually* blacklisted:
The website returns that operation for any extensions for which it
doesn't find any versions that match the shell version. That is, the
most likely reason is that the user updated to a new GNOME release
which the extension doesn't support yet.
It doesn't look like the website is going to change that behavior any
time soon[0], so drop the 'blacklist' handling for the time being.
Florian Müllner [Thu, 26 Mar 2020 22:40:38 +0000 (23:40 +0100)]
js: Add missing chain-ups in vfuncs
Commit 55b57421d changed signal handlers to the corresponding vfuncs,
but didn't always chain up as necessary. In most places this doesn't
matter, but at the very least the commit broke activating message list
items via the keyboard.
Add all (hopefully) the missing chain-ups to get the expected behavior
back.
Florian Müllner [Mon, 23 Mar 2020 19:59:31 +0000 (20:59 +0100)]
messageTray: Don't create notification policy on demand
This was changed in commit 8f15193b4 as a work-around for an ES6
class limitation, but now that Sources are GObject subclasses, we
no longer need to use that somewhat surprising pattern.
Florian Müllner [Wed, 25 Mar 2020 22:49:58 +0000 (23:49 +0100)]
messageTray: Only destroy policy after emitting ::destroy
Destroying the policy invalidates it, so accessing it from a
Source::destroy handler (for example to disconnect signal
handlers) currently results in warnings like:
Object .Gjs_ui_messageTray_NotificationApplicationPolicy
(0x7f8c7c0a64a0), has been already deallocated — impossible
to access it.
Florian Müllner [Wed, 25 Mar 2020 15:44:26 +0000 (16:44 +0100)]
st/icon: Add a default fallback icon
The idea behind commit 3dd8ffc2bb1cefd to try harder to avoid empty
icon actors because of missing icons was sound, so implement that
behavior in StIcon itself:
If the main gicon was not found, and the fallback gicon isn't set or
wasn't found either, fall back to the standard 'missing-image' icon.
Florian Müllner [Wed, 25 Mar 2020 15:40:47 +0000 (16:40 +0100)]
Revert "texture-cache: Use image-missing image when no other icon was loaded"
The commit broke StIcon's :fallback-gicon property, as it relies on failure to load
an icon to determine that the fallback should be shown.
Luckily StIcon is the only user of st_texture_cache_load_gicon() (at least in
regular shell code), so we'll be able to implement the 'image-missing' fallback
there.
Florian Müllner [Thu, 5 Mar 2020 17:01:38 +0000 (18:01 +0100)]
extensionPrefs: Export parent window for prefs dialog
Now that OpenExtensionPrefs()'s parentWindow parameter is actually
supported, export the main application window so it can act as
transient parent to the prefs dialog.
Now that the service implements the preference dialog, it's time
to support OpenExtensionPrefs()'s parentWindow parameter and make
the dialog a transient of the external window if specified.
Florian Müllner [Thu, 5 Mar 2020 03:52:09 +0000 (04:52 +0100)]
shew: Add small library for dealing with external windows
In order to support OpenExtensionPrefs()'s parentWindow parameter,
we will need the ability to make a window transient to an external
window identified by a string handle.
This takes the corresponding code from xdg-desktop-portal-gtk and
brings it into an introspectable form, likewise the counterpart in
libportal to export a string handle for a GtkWindow.
Since PackageKit 1.11.1, the prompt to install updates on the end
session dialog has been (mostly) broken. The problem is that it only
works if PackageKit is running at the time the end session dialog is
opened; otherwise, our GDBusProxy has invalidated all of its properties,
which we read to see if update is possible. We need to autostart
PackageKit before reading its properties to fix this problem. That would
be easy if we were calling a method to see if an update or distro
upgrade were available, but since we're just checking a property, using
cached properties won't suffice. We'll have to manually check the
property value to ensure we autostart PackageKit.
Most of the code is written by Florian. Thanks Florian!
Daniel van Vugt [Tue, 17 Mar 2020 07:58:21 +0000 (15:58 +0800)]
layout: Show system background and animate on the same frame
Previously we'd show the system background and then wait till the
main loop was idle before beginning the shell startup animation.
This resulted in one initial frame that was always just the system
background.
Now we try to get both the system background and the startup animation
begun on the same first frame.
telepathyClient: Use proper Object to wrap different tpl messages
In telepathyClient we consider messages both Tpl.TextEvents and
Tpl.Messages, and we manually create JS objects to copy the properties we
care for each one. This may lead to objects not matching the interface we
want.
Instead, use an object with construct-only properties and two factory static
methods to initialize it.
Unfortunately we need to use the ChatMessageClass for the class name or
calling the static methods would trigger a gjs error as per [1].
Ray Strode [Wed, 25 Mar 2020 20:30:46 +0000 (16:30 -0400)]
data: ensure systemd environment is sanitized when shell exits
When mutter is acting as a display server it sets a number of
environment variables in the user's session. These variables
tell applications where the display server's sockets are.
When the shell exits at logout time it leaves these environment
variables in the systemd --user environment, which can confuse
subsequent sessions.
telepathyClient: Use GObjects based message objects
As per commit b5676a2a5 ChatNotification is a GObject, but I was wrongly
considering that it was using Tp.Message's as children, instead it just
uses custom-built objects to pass information around through signals.
Given gjs can only use GObjects as signal parameters, create a small wrapper
class to hold the ChatNotification messages and use it as signal parameter.
Florian Müllner [Fri, 20 Mar 2020 11:42:04 +0000 (12:42 +0100)]
main: Do not warn about missing GDM on each login
We now warn on startup if screen locking isn't available, however for
users who choose not to use GDM or logind, repeating the warning on
each login is more annoying than helpful.
Instead, limit the warning to the first login on which the screen lock
became unavailable. That way the notification will still serve the
intended purpose of informing the user, but without being perceived
as nagging.
Florian Müllner [Mon, 23 Mar 2020 17:00:27 +0000 (18:00 +0100)]
environment: Hook up touch_file to GFile prototype
We don't usually extend introspected types with our own API, but in
this case it's too tempting to make the helper functions usable with
Gio._promisify() ...
Florian Müllner [Sun, 22 Mar 2020 17:00:00 +0000 (18:00 +0100)]
extensionPrefs: Ensure up-to-date release version in metainfo
It's easy to forget to add a new <release> tag to the metainfo when
doing a new release.
Address this with an additional test if appstream-util is recent
enough to include the new validate-version command, so distcheck
fails when the metainfo wasn't updated.
Florian Müllner [Tue, 24 Mar 2020 06:15:26 +0000 (07:15 +0100)]
shell/screenshot: Throw error on failure
Commit da537cda43 moved the Shell.Screenshot API to GIO's async pattern,
but we never set the GError passed to the *_finish() functions and only
indicate failure by returning FALSE.
The expected behavior is to throw an error in that situation, so make sure
we do that.
Florian Müllner [Tue, 24 Mar 2020 06:10:27 +0000 (07:10 +0100)]
shell/screenshot: Remove unnecessary NULL check
Since commit be5f5ec9d40, the output stream is created externally and
expected to be non-NULL, so the check is now pointless (in particular
*after* calling g_object_ref() on the stream).
Florian Müllner [Mon, 23 Mar 2020 22:22:17 +0000 (23:22 +0100)]
ibusManager: Simplify code a bit
ibus_bus_request_name_async_finish() will throw an error on failure,
so we can move the error handling there instead of checking for the
return value, which saves one level of indentation.
texture-cache: Use image-missing image when no other icon was loaded
If a given icon is not available for neither the current theme nor any
of the fallback options (default theme -"Adwaita", "gnome"...), the
call to gtk_icon_theme_lookup_by_gicon() will return a NULL value, which
returned by the Shell Toolkit as is, causing trouble in the Shell's JS
code when calling shell_app_create_icon_texture() to create an icon.
To at least mitigate the chances of this having this issue happening, we
should at least try to load the standad 'image-missing' icon from the
Icon Naming Specification spec when we receive a NULL here, so that
at least we try to show something to the user, even if it's ugly.
Florian Müllner [Sat, 21 Mar 2020 01:32:08 +0000 (02:32 +0100)]
extensions-tool: Use new Extensions proxy
While sandboxing isn't a concern for the gnome-extensions command line
tool, using the Extensions proxy directly means that D-Bus methods are
called from the tool rather than gnome-shell, which allows the proxy
to auto-shutdown when done.
Florian Müllner [Wed, 4 Mar 2020 04:05:08 +0000 (05:05 +0100)]
extensionPrefs: Use new Extensions proxy
We will soon sandbox the application, so it makes sense to not request
more access than strictly necessary (even with priviledged access like
extension management).
Florian Müllner [Wed, 4 Mar 2020 04:05:08 +0000 (05:05 +0100)]
dbusServices/extensions: Take over prefs dialog from app
As outlined earlier, in order to turn the Extensions app into a properly
sandboxed application, we need to split out the extension prefs dialog
and move it elsewhere.
With "elsewhere" being the new Extensions D-Bus service, effectively
turning it into a shell extensions portal.
Florian Müllner [Wed, 4 Mar 2020 03:02:28 +0000 (04:02 +0100)]
dbusServices/extensions: Proxy Extensions API
Similar to the previously added org.freedesktop.Notifications proxy,
this exposes the org.gnome.Shell.Extensions API and forwards any
request to the real implementation in gnome-shell.
The motivation differs though: We want to be able to package the
extension app as flatpak and distribute it separately, but the
extension prefs dialog is hard to impossible to sandbox:
- filenames need translating between host and sandbox, and we
can only do that in some cases (serializing/deserializing
extensions), but not others (extension settings that refer
to files)
- system extensions install their GSettings schemas in the system
path; the best we can do there is assume a host prefix of /usr
and set GSETTINGS_SCHEMA_DIR in the flatpak (eeks)
- extensions may rely on additional typelibs that are present on
the host (for example because gnome-shell itself depends on
them), but not inside the sandbox - unless we bundle all of
gnome-shell's dependencies
- if gjs/mozjs differ between host and sandbox, extensions must
handle different runtimes for the extension and its prefs
And all those issues occur despite a very permissive sandbox (full
host filesystem access, full dconf access, full org.gnome.Shell
access (including Eval()!)).
This new service will give us an alternative place for handling
the preference dialog:
- it runs outside of gnome-shell process, so can open windows
- it runs on the host, so the extension's prefs get to run
in the same namespace as the extension itself
That is, the service will provide portal-like functionality (albeit
not using the org.freedesktop.portal.* namespace, as extension
management is an inherently privileged operation).
Florian Müllner [Thu, 5 Mar 2020 18:44:28 +0000 (19:44 +0100)]
extensionPrefs: Stop handling UUIDs on the command line
We are jumping through quite some hoops to support showing only the
preference dialog when given a UUID on the command line.
As gnome-shell is about to stop calling out to us for the prefs dialog,
the reason for supporting this is going away, so remove all the special
handling.
projects / gnome-shell.git / log