LayoutManager is currently a pure JavaScript class that
relies on the rudimentary Signals.addSignalMethods() to
handle signals. This is an inefficient implementation of
one of the most central classes in GNOME Shell.
In addition to removing Shell.GenericContainer, then,
turn LayoutManager into a proper GObject.Object subclass.
This one was remarkably easy to port. In order to make it,
replace the Shell.GenericContainer handlers by a constraint
and simply replace it by a St.Widget.
In the next commit, we will turn PanelMenu.ButtonBox into a
St.Widget subclass. As a domino effect, PanelMenu.Button will
become one too, and so will Panel.AppMenuButton.
When that happens, the current show() and hide() functions in
Panel.AppMenuButton will clash with Clutter.Actor's ones.
To avoid that, rename these functions to fadeIn() and fadeOut()
and avoid a name clash.
This is another straight port from Shell.GenericContainer.
The important thing to notice is that the calculation is
broken if the StThemeNode helpers (adjust_preferred_* and
adjust_for_*) aren't used.
The downside of this patch is that it removed the skip_paint
from the thumbnails. Keeping it would add an unecessarily
large amount of code.
Removing Shell.GenericContainer from the IconGrid class was
challenging because it needs the "skip paint" API from it.
This API was added, too, as a workaround to the inability
to override vfuncs from GJS.
The overrides are largely copy-pasted and translated versions
of the Shell.GenericContainer code.
The IconGrid:key-focus-in signal was renamed to :child-focused
to avoid clashing with ClutterActor:key-focus-in.
In GridSearchResults, the internal IconGrid had it's y_expand
set to false, so it doesn't push other search elements (the
list results mainly) to the bottom of the screen.
Because skip paint wasn't and still isn't a GObject property,
rename it to _skipPaint to reflect that.
As part of our quest to obsolete Shell.GenericContainer, IconGrid will
become a Clutter.Actor subclass. As the ::key-focus-in signal would
clash with Clutter.Actor::key-focus-in, rename it to ::child-focused.
DashItemContainer currently animates the scale and opacity
of its child when zooming in. This is visible when adding
a new favorite item to the dash; the items will zoom in from
the center.
After the previous commit, however, the zoom animation got
slightly broken, and looked like the icon was coming from
the bottom instead of the center.
Fix that by setting the scale and opacity of DashItemContainer
itself, instead of its child. Remove the unused code after that
too.
Pretty much like the previous patches, this extends St.Bin. The
most interesting aspect of this patch is that most of the sizing
routines of the icons is now delegated to the actors and layout
managers, removing quite a bunch of code.
The 'spacing' theme property is now redirected to StBoxLayout's
spacing property. Also adjust the Dash code to stop forcing a
potentially invalid width in the first icon too.
st-box-layout: Pass correct allocation box to layout manager
StBoxLayout implements StScrollable, which, semantically, means that
the StBoxLayout size may not match the minimum size reported by the
layout manager. In this specific case, the layout manager by is a
ClutterBoxLayout by default. For example:
- ContentSize = the minimum size of the content;
- ViewportSize = the allocated size of the viewport;
When allocating StBoxLayout, it must assume ViewportSize, but must
pass ContentSize to the layout manager. That way, the children of
StBoxLayout are correctly placed within it, even if it's bigger than
ViewportSize.
And here's the problem: right now, StBoxLayout assumes ViewportSize
AND also passes it to layout manager. Commit 77c4c6b6d specifically
exposed this bug by relying entirely on StBoxLayout to arrange the
app and window icons.
Fix that by using ViewportSize to allocate StBoxLayout itself, but
passing ContentSize to the layout manager.
This commit removes all the uses of Shell.GenericContainer from
SwitcherPopup.SwitcherList. Compared to the other patches, this
one was specially trickier to get right, and a few invasive
changes needed to be done.
The most noticeable one is that the allocation of the items is
done entirely by St.BoxLayout -- we don't manually allocate them
anymore. To make it work, get_preferred_width() had to calculate
the correct value. It now assumes that:
* Minimum width: the minimum width of the widest child.
* Natural width: the minimum width of the StBoxLayout (use it
instead of the natural width to force the labels to ellipsize
when too long.)
The AppIcon class became a St.Widget subclass as well, to override
get_preferred_width() and be able to keep the squared shape.
Besides that, add a new SwitcherButton class to reimplement squared
icons without having to resort to hacks in the size allocation
machinery. This class has a single vfunc override to ensure that it
is squared when the SwitcherList is.
The arrows indicating multiple windows are now in this._list
actor to the SwitcherPopup itself, since this._list automatically
manages its own children now.
At last, adapt (but preserve) the hack in CyclerPopup.
switcherPopup: Use MonitorConstraint instead of vfunc overrides
Instead of overriding vfunc_get_preferred_width|height(), use the
already available Layout.MonitorConstraint to bind SwitcherPopup
to the primary monitor.
This commit turns SwitcherPopup.SwitcherPopup into a St.Widget
subclass, and gets rid of Shell.GenericContainer usage. Subclasses
were adapted to that too.
This class introduced a new challenge: it overrides show(). As per
discussions, we now call this.visible = true inside show().
switcherPopup: Rename destroy() to fadeAndDestroy()
In the process of purging all usages of Shell.GenericContainer
of GNOME Shell, one specific problematic situation that might
occur is when classes have functions that would clash with any
ClutterActor or StWidget function name.
One of such example is SwitcherPopup.destroy(). Right now, this
class is a pure JavaScript class that wraps a real actor, but
soon this will change, and it'll become a St.Widget subclass.
Another problem with functions that mimic the toolkit ones is
the predictability of them; after calling destroy(), that widget
is expected to not be available anymore. In SwitcherPopup case,
it is still available for a short while. In this case, that's not
a big problem, but the show() and hide() functions in other clases
are more problematic because the actor's visibility does not
follow that.
This commit is a first step in cleaning that up, and changes the
SwitcherPopup.destroy() to fadeAndDestroy().
Shell.GenericContainer exposes the size negotiation machinery
through the use of signals. Signals are not specially performant.
One of the reasons is that they acquire a global lock for signal
handlers lookup. GNOME Shell has more than 2,000 actors at any
given point in time, up to 20 levels deep in hierarchy, making
size negotiation and painting non-trivial tasks. Such a critical
section of Clutter's machinery shouldn't rely on signals
whatsoever.
Regardless of that, Shell.GenericContainer is a workaround to
a non-existing issue anymore. It shouldn't be used anyway, and
any performance improvements that removing it can potentially
yield are bonuses to it.
This commit starts this work by removing Shell.GenericContainer
usage from Panel.Panel class. The class now extends St.Widget,
and as such, it has no "this.actor" field set anymore. A couple
of places where this actor field was used are adjuste as well.
It is important to notice that we now allocate the Panel itself
inside vfunc_allocate(). This was previously done before emitting
the signal by Shell.GenericContainer.
Carlos Garnacho [Thu, 27 Sep 2018 19:09:02 +0000 (21:09 +0200)]
inputMethod: Use forward_key() method to forward key events
ClutterVirtualInputDevice has the limitation that event flags won't be
made to contain CLUTTER_EVENT_FLAG_INPUT_METHOD, possibly causing feedback
loops.
As the event gets injected up the platform dependent bits, we can avoid
care on not pressing the same key twice, we still expect coherence between
key presses and releases from the IM though.
Florian Müllner [Sun, 12 Aug 2018 00:17:42 +0000 (02:17 +0200)]
appFavorites: Don't add app explicitly
When adding a favorite, we add the ID to the list of favorites, save the
setting and add the new app to the favorites map. However as writing the
settings value already results in reload() to update the favorites map,
the new app is usually already in the map when we add it.
The only exception is when the ID was found in the RENAMED_DESKTOP_IDS map,
in which case we end up adding both the renamed app and the original one.
Fix this by simply relying on reload() to properly update the map, just like
we already do in _removeFavorite().
closeDialog: Untrack chrome when window loses focus
On X11, reactive chrome must be added to the input region in order
to work as expected. However that region works independently from
any window stacking, with the result that the unresponsive-app dialog
currently blocks all input in the "covered" area, even in windows
stacked above the unresponsive window.
The correct fix would be to track the unobscured parts of the dialog
and set the input region from that, but that's quite cumbersome. So
instead, only track chrome when the corresponding window is focused
(or the dialog itself of course).
Florian Müllner [Fri, 5 Oct 2018 17:09:28 +0000 (19:09 +0200)]
automountManager: Explicitly track active operations
As a mount operation's UI may be reused (for example after mistyping
the password), we only close the operation once the mount has finished
(successfully or with error).
We therefore need to track ongoing operations, which we currently do
by monkey-patching the corresponding volume object. However while the
underlying GVolume object indeed remains the same through-out the
operation, the JS wrapper object isn't referenced anywhere and may
thus be garbage collected, resulting in a stuck dialog.
Fix this issue by tracking active operations explicitly, so that all
involved objects are referenced until the end of the operation.
Whenever a command runs in the run dialog, it
will be added to the history unless it is
already the last entry. This does not apply
for entries that are not consecutive, which can
result in long chains of commands which
alternate, e.g. lg, r, lg, r, lg, r. Not only is
this wasteful in terms of space, but also
inconsistent with how history works elsewhere,
e.g. in the shell.
Therefore, remove entries in the history that are
equal to the one that will be added to the end of
of the history when the entry already exists.
networkAgent: Ask for wifi secrets in the hints paremeter
The `hints` and `settingName` parameters to the agent call may define
the specific list of secrets NM actually needs from the user. This
seems to have been the intended use of these two parameters but only
recently did NM with the IWD backend start to use this to request 802.1x
secrets. So if `hints` is provided, ask user for the specific secrets
listed there and don't even look at what type of EAP method is in use.
Only the three types of secrets actually in use by NM's IWD backend are
supported for now -- they happen to be the same three that
_get8021xSecrets() had already supported.
gjs now relies entirely on introspection data to determine parent
types and implemented interfaces, so in order to have all methods
and properties resolve correctly, we must include the corresponding
GIRs of all types used.
This attribute was previously only assigned in show(). hide() compares
this attribute to 0. If hide() is called before show() is first called,
the comparison would give the correct result (undefined > 0 is false)
but log a warning:
JS WARNING: [resource:///org/gnome/shell/ui/workspacesView.js 529]:
reference to undefined property "_restackedNotifyId"
Initialize this attribute in _init(), alongside _scrollEventId and
_keyPressEventId which are also used in hide().
Will Thompson [Tue, 25 Sep 2018 17:01:36 +0000 (18:01 +0100)]
endSessionDialog: squash "reference to undefined property" warning
dialogContent is set to one of the elements of the list DialogContent,
but not all of those have a checkBoxText property. When logging out (as
opposed to shutting down), this causes a warning:
JS WARNING: [resource:///org/gnome/shell/ui/endSessionDialog.js
763]: reference to undefined property "checkBoxText"
(The line number corresponds to this line in 3.28.3.)
The warning is apparently not triggered if the undefined property is
used as part of a boolean expression:
gjs> var x = {};
gjs> x.a;
typein:2:1 strict warning: reference to undefined property "a"
gjs> if (x.b) { log('oh no'); }
gjs> x.c || ''
""
_setCheckBoxLabel() just checks the truthiness of its 'text' argument,
and the empty string is false-y, so passing '' rather than undefined has
no functional effect.
Adam Williamson [Tue, 18 Sep 2018 23:05:42 +0000 (16:05 -0700)]
Fix connection to wifi APs from user menu (RH #1628263)
In recent Fedora 29, connecting to wifi access points from the
user menu (top-right menu) does not work. Clicking the 'Connect'
button just animates it but does nothing else. The logs show an
error "JS ERROR: Error: Expected type utf8 for Argument
'specific_object' but got type 'undefined'".
Looking into this, it seems the problem is these uses of the
`path` property of an NMAccessPoint. NMAccessPoint inherits
from NMObject, and NMObject *does* have a path property:
so at first glance this seems fine. But I poked around a bit
using libnm via Python (which goes via introspection, just like
this JS code does), and found that indeed AccessPoint objects
don't seem to have a `path` property there either.
Looking at the libnm code, this actually makes sense, because
the property is marked "(skip)":
I'm a bit concerned that this was only found recently - whereas
the change to use `.path` in gnome-shell dates from October 2017
(d71af5e5) and the property has been marked (skip) in NM since
at least 2016 - but this all seems to add up. The obvious fix is
to replace use of `.path` with `.get_path()`, which returns the
path and is *not* marked (skip) and so *is* available via
introspection. I tested that this works in Python and also did
a test build of gnome-shell with this change and installed it on
an affected system, it does seem to fix the bug.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Commit a76b28bd moved the dbus-interfaces to the
dbus-interfaces directory without changing the
path for docs build as well, resulting in a fail
of the build when building with gtk_doc=true.
So change the path in the build accordingly to
ensure that the build does not fail when building
with gtk_doc=true.
gjs's D-Bus convenience explicitly expects a string representation
of an interface, but the new convenience method to load an XML
description from a resource introduced in commit f42d9df3e0 only
returns a string when using gjs from the GNOME 3.30 release. We
have so far managed to keep compatibility with the previous stable
gjs release, so fix up the fallback code to cast to string.
fileUtils: Add helper for loading D-Bus XML from resource
Commit dbf993300a moved all inline D-Bus interface descriptions to template
strings so we can stop escaping line breaks.
Unfortunately that unveiled a grave bug in xgettext, which currently cannot
handle files that contain both backtick and slash characters - as a result,
translations from affected files have started to disappear as translators
run xgettext/msgmerge.
Instead of reverting the change and getting the crusty escaping back, we
will take this as an opportunity to stop inlining the XML altogether and
load it from a resource instead.
To facilitate that, add a small helper method that loads a D-Bus interface
description from a dedicated resource bundle.
It is used by at least Software, Tweaks and the webextension, so it's
a bit odd to install documentation for lesser used interfaces but not
this one ...
data: Move D-Bus interface descriptions into a subdir
We'll soon move the inlined XML of all interfaces we export or consume into
separate files. That's a significant increase from the interfaces for which
we install documentation, so it makes sense to use a dedicated subdir.
Daniel Drake [Tue, 31 Jul 2018 00:17:44 +0000 (21:17 -0300)]
switchMonitor: switch to next config upon initial keypress
In GNOME-3.24, pressing Super+P or a similar function key would cause
a switch to the next available monitor configuration.
However, in GNOME-3.26, this was reimplemented in mutter and gnome-shell
and the behaviour is now different: pressing Super+P and releasing will
cause no change in montor configuration[1]. In this new design you have
to press Super+P and keep holding Super in order to keep the switcher
open, then press P again (or use the arrow keys or mouse) to
select the next one in the list.
This is incompatible with many Asus products such as Asus X530UN, where
pressing the presentation mode media key (Fn+F8) actually generates
the following keypress events from the keyboard controller:
Fn pressed: nothing
F8 pressed: nothing
F8 released: Super press, p press, p release, Super release (quick burst)
Fn released: nothing
With this firmware behaviour it's not possible to hold the keys and have
the dialog come up so that you can select another new mode.
To solve this, when the switcher is opened, select the next available
display config by default, which is more similar to the pre-GNOME-3.26
behaviour. Now pressing Fn+F8 on this laptop will result in the display
mode switch taking place.
[1]: The mentioned desired behaviour will at least happen after
https://gitlab.gnome.org/GNOME/mutter/issues/281 has been fixed
Since we always keep the active workspace until the user switches
to a different one, we may end up with two empty workspaces at
the end. It's not obvious to users why this happens, and there's
indeed no good reason for the behavior - just remove the trailing
workspace in that case.
There's no relation between a window being hidden from overview/taskbars
and a window not being closable - currently we effectively disable the
fallback quit action for any application with open transients, which
simply doesn't make sense.
Instead, only exclude windows for which the close action has been
explicitly disabled.
Iain Lane [Thu, 16 Aug 2018 10:59:36 +0000 (11:59 +0100)]
network: Don't assume the active connection has been processed first
`NMConnectionDevice._sync()` is responsible for setting up the active
connection that we'll end up displaying. It expects the active
connection to already be in a map `_connectionItems`. If it isn't in
there, we get a null dereference and the indicator can get into a weird
state where it doesn't display devices / connections properly.
Let's change this expectation. If there is an active connection,
`_deviceAdded()` will eventually get to it and call `_sync()` to set up
the active connection state. We make `_sync()` tolerate there being no
active connection when it's called.
extensionSystem: Unset stylesheet file reference when unloaded
We must remove the GFile reference from the representing object when an
extension has been unloaded as this won't be used anymore later (e.g. as cached
ref).
js/main: Throw error if no valid default stylesheet is found
Throw an error using an informative message in case a mode uses a stylesheet
that can't be loaded, instead of crashing later because the theming can't be
properly computed, and thus the minimum size of the actors.
extensionSystem: Don't load an extension with invalid stylesheet
We currently assign the stylesheet to an extension whenever the file exists,
regardless of whether it actually loaded successfully or not.
And thus we load an extension that ships a stylesheet even if that file can't
be used.
There is no point in trying to load an extension if its stylesheet wasn't
loaded in the first place, so make sure this happens only on success.
search: Cancel search provider operations on clear
Ensure that the search provider operations (just getResultMetas requests
in the current implementation) in progress are properly cancelled when we
clear the UI, otherwise returned results might still be added when not
needed.
This is triggered for each provider by the SearchResults reset.
Currently when the overview is hidden, any pending search is kept alive,
not only at remote search provider level (as per issue #183), but even
the shell providers proxies continue to get and process data. This happens
even if this is not needed anymore, while the UI reset is performed only
next time that the overview is shown (causing some more computation
presentation time).
In order to stop this to happen, when the overview is hidden, we have to
unset the search entry to an empty value as this would make SearchResults
to have empty terms list and that would make the proxies cancellable to
be triggered (without causing any further search to start).
automountManager: remove allowAutorun expire timeout on volume removal
If the volume is removed before AUTORUN_EXPIRE_TIMEOUT_SECS seconds, we can stop
the timeout earlier as there's nothing to unset, while the volume instance
won't be valid anymore.
messageList: stop syncing if closeButton has been destroyed
The _sync function for Message only updates the close button visibility,
so we can safely stop doing that if the close button get get destroyed earlier
(as it happens when clicking on it).
dnd: Nullify _dragActor after we've destroyed it, and avoid invalid access
We need to avoid that we use the _dragActor instance after that it has
been destroyed or we'll get errors. We now set it to null when this
happens, protecting any access to that.
Add a DragState enum-like object to keep track of the state
instead of using booleans.
Remove duplicated handler on 'destroy' and just use a generic one.
Ray Strode [Fri, 31 Aug 2018 16:48:00 +0000 (12:48 -0400)]
objectManager: correct invalid index code in onNameVanished
The object manager tries to synthesize interface removal
events if the bus name of a remote object drops off the bus.
The code has a bad typo in it, though: it confuses `objectPaths`
(the list of all object paths) and `objectPath` (the object
currently being processed this iteration of the loop).
That leads to a failure to synthesize the interface removal
events, and spew in the log.
This commit corrects the objectPath/objectPaths confusion.
projects / gnome-shell.git / log