Currently lookingGlass relies on some odd BoxLayout behavior, which
allows children to use fixed positioning without affecting the parent's
size request. As this behavior is scheduled for removal, add the
looking glass dialog directly to Main.uiGroup.
https://bugzilla.gnome.org/show_bug.cgi?id=703808
The duality of the Clutter's key focus and mutter's window focus has long been
a problem for us in lots of case, and caused us to create large and complicated
hacks to get around the issue, including GrabHelper's focus grab model.
Instead of doing this, tie basic focus management into the core of gnome-shell,
instead of requiring complex "application-level" management to get it done
right.
Do this by making sure that only one of an actor or window can be focused at
the same time, and apply the appropriate logic to drop one or the other,
reactively.
Modals are considered a special case, as we grab all keyboard events, but at
the X level, the client window still has focus. Make sure to not do any input
synchronization when we have a modal.
At the same time, remove the FOCUSED input mode, as it's no longer necessary.
https://bugzilla.gnome.org/show_bug.cgi?id=700735
We can easily implement much of the same behavior ourselves by
keeping track of Clutter's focus events. Reintroduce heavily
modified FocusGrabber to do the work for us.
This will temporarily break when the user selects a window until
we can make gnome-shell automatically set the stage focus.
This also removes our only use of focus grabs, so remove those
as well.
https://bugzilla.gnome.org/show_bug.cgi?id=700735
We can't assume "isActive implies isModal", so there is a risk
of pushing a modal that nothing else will ever pop, because we
take the early return and don't activate the user active watch.
https://bugzilla.gnome.org/show_bug.cgi?id=700901
_updateIcon should not attempt to sync any active connections, as the
icon-changed signal can be emitted in response to something done during
_syncActiveConnection. In the case of VPN, removeActiveConnection would
cause an icon-changed signal to be emitted immediately, but the state
would not be updated, causing us to call removeActiveConnection over and
over.
Explicitly sync all active connections when we know it needs to be done,
and simply make _updateIcon synchronize with the current device's icon.
https://bugzilla.gnome.org/show_bug.cgi?id=703565
This fixes a blue background being drawn when switching the monitors
configuration using hardware keys
(clone/multimonitor/external/internal).
The problem is that the shell gather all background loading requests
under the same meta_background_load_file_async call using one
GCancellable (the first one to come). So when the shell receives a
batch of 12 or so XRandr events, it creates 12 new background managers
which end up trying to load 12 times the same background picture. All
of these requests are batched into the same
meta_background_load_file_async using the first GCancellable received
on the first request. Unfortunately, when the first request is
cancelled by the following event indicating a new monitor setup, all
of the background picture requests are dropped on the floor, and
nothing gets loaded (hence the blue screen background).
https://bugzilla.gnome.org/show_bug.cgi?id=703001
It turns out that picking a 3200x1200 scene on notebook chipsets
every time the mouse is moved isn't exactly the fastest thing. Defer
picking to an idle to ensure that it won't get in the way of keeping
up with mouse events.
https://bugzilla.gnome.org/show_bug.cgi?id=703443
The code that checks for various conditions is confusing and
undercommented. It appears one of the recent refactorings
inadvertedly inverted the sense of the 'hidden mountpoint'
check, and caused autorun to not work for anything that does
not have a 'native root' - which is pretty much all volumes
implemented by gvfs.
https://bugzilla.gnome.org/show_bug.cgi?id=703418
Commit d6cace32 introduced a typo in the left/right arrow side
calculation code that causes in most scenarios (where the monitor
width is greater then the height) to not flip the box when it doesn't
fit inside the monitor.
https://bugzilla.gnome.org/show_bug.cgi?id=703403
If the drag action ends after something else has put the screen shield
into a different state we can end up in an inconsistent screen shield
state where the whole thing is empty.
https://bugzilla.gnome.org/show_bug.cgi?id=703126
Commit 16fa186b63 attempted to fix the zoom animation problem
by throwing changes on the floor while the overview is animating. This has
the side effect that we might end up missing some positioning changes causes
windows to overlap the workspace thumbnails.
So revert those changes and fix it by simply by passing
WindowPositionFlags.ANIMATE during the overview animation.
This way the animation works as expected and we don't miss any position changes.
https://bugzilla.gnome.org/show_bug.cgi?id=703105
The top_window_group was introduced for popup windows that should
appear above system chrome, but as the group itself is just a child
of Main.uiGroup, chrome that is added after top_window_group will
still be stacked on top.
At least correct the stacking for actors added via addChrome().
https://bugzilla.gnome.org/show_bug.cgi?id=702338
The class is generally useful, so it only makes sense in panel.js
for historical reasons. Because other parts of the code are
using it, though, problems are cropping up that require a
workaround like:
placeSpinner: function(...) {
/* This is here because of recursive imports */
const Panel = imports.ui.panel;
Panel.AnimatedIcon(spinnerIcon, WORK_SPINNER_ICON_SIZE);
...
}
This commit moves AnimatedIcon to its own file so we can drop that
workaround.
https://bugzilla.gnome.org/show_bug.cgi?id=702818
When the osd window is hidden based on the timeout, it accidentally
left the timeout ID in place. When a subsequent switcher popup came
up, it thought the OSD window was scheduled to be hidden and tried
to re-hide the actor. This caused the tween to be run along with
an extra call to enable_unredirect_for_screen.
When the allocation of the workspacesView changes during the animation we override
the tween with one that does not animate causing the overview zoom animation
not to happen.
Fix that by ignoring the alloactionChanged notification during the overview
animation.
The code here before was trying to play hierarchy tricks to
figure out how to show / hide the events list, which broke
when we rearranged how the date menu was laid out. Simplify
the code here to not be so tricky, and update the CSS to
match the new designs.
https://bugzilla.gnome.org/show_bug.cgi?id=702849
The timestamp timeout specifies how long we should wait before
adding a timestamp to the notification. A timeout of one minute
ended up showing a lot of timestamps, so increase it to 3 minutes.
https://bugzilla.gnome.org/show_bug.cgi?id=687809
If we focus notifications before they're expanded, the body and action
area won't be visible, and the can_focus members like the text entry
will not be able to be focused.
Ensure that all of the all actors that would be in an expanded notification
are visible before we attempt to focus them.
https://bugzilla.gnome.org/show_bug.cgi?id=698778
Migration from old settings can result in a path instead of URI
there. This is technically invalid, but can easily recognize it
and avoid the crash.
Minor changes by Ray Strode
https://bugzilla.gnome.org/show_bug.cgi?id=702121
The status item will go away soon, so make sure the one-time
fire is given its own function. At the same time, only connect
to the signal when the situation actually matters.
https://bugzilla.gnome.org/show_bug.cgi?id=701954
No class in here has this.carrier as a property. Presumably, this was
meant to be this.device.carrier, but since this code is going to be
rewritten soon anyway, might as well just junk the never-working
code for now.
https://bugzilla.gnome.org/show_bug.cgi?id=701954
In order to have event descriptions on multiple lines, but still
maintain proper alignment with the day and time strings, refactor
the whole event list to be one big table. Headers are implemented
as spanning cells, and uneven spacing is a mix of row/column spacing
and cell padding.
https://bugzilla.gnome.org/show_bug.cgi?id=701231
In BlueZ 4, Authorize() was used to authorize both service
and JustWorks authorization requests. In BlueZ 5 these two
have been split into AuthorizeService() for services and
RequestAuthorization for JustWorks devices. Adapt the
Bluetooth code accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=700891
This can be more easily achieved by listening for changes to the
device's active-connection property. VPN will still need support to
track active connections, as it does not have an associated
device. But as VPN can track multiple active connections, the names
"set" and "clear" don't quite fit. Rename them to the more-standard
"add" and "remove".
https://bugzilla.gnome.org/show_bug.cgi?id=701954
The end session dialog was waiting a second before updating
its text to display the timer. It is nicer to show the correct
message from the start.
https://bugzilla.gnome.org/show_bug.cgi?id=702056
We currently monitor the shell's override schema for changes to
the 'dynamic-workspaces' key, which ends up being the wrong
schema in classic mode. With the new ability to use mode-specific
overides, we can finally fix this.
https://bugzilla.gnome.org/show_bug.cgi?id=701717
This will allow the use of mode-specific defaults. For classic mode
we currently implement this with mini-extensions, but this may result
in confusing behavior when settings change due to extensions being
disabled during screen locks (not to mention that those mini-extensions
are hardly an elegant approach).
https://bugzilla.gnome.org/show_bug.cgi?id=701717
We will allow to use mode-specific overrides; in preparation for that,
move the code so that we only override preferences after initializing
the session mode.
https://bugzilla.gnome.org/show_bug.cgi?id=701717
As multiple-connections for a Wi-Fi AP won't fit in the new design,
remove submenus right now. Simply make a simple item that connects
to the first known connection for the AP, which should be the common
case.
https://bugzilla.gnome.org/show_bug.cgi?id=698918
Currently we "only" grab the keyboard when starting a drag operation,
which does not impede keybindings to be processed. This is at best
not harmful (like workspace switching), but may have unintended effects
otherwise - for instance, the hot corner is disabled, so having the
corresponding keyboard shortcut still active is fairly odd (not to
mention that it leaves the system in a confused state).
Fix this by switching to pushModal()/popModal(), which will push a
dedicated keybinding mode for us.
https://bugzilla.gnome.org/show_bug.cgi?id=700877
Currently the clipboard's contents may leak to unauthorized parties by
pasting into the unlock dialog's password entry and unmasking the entry.
Prevent this from happening by clearing the clipboard on lock.
https://bugzilla.gnome.org/show_bug.cgi?id=698922
When the dash does not contain any applications (either favorites
or running), it is currently impossable to add a favorite via DND.
Grow the dash slightly in that case to provide a drop target.
https://bugzilla.gnome.org/show_bug.cgi?id=684618
We currently only keep track of old placeholders when moving past
the dragged app's current favorite position, as this is the only
case where we need to worry about jitter. Still, moving it into
_clearDragPlaceholder() allows us to consolidate code paths, which
is a good thing ...
https://bugzilla.gnome.org/show_bug.cgi?id=684618
The function currently only resets the placeholder position if
there is a placeholder; this is not necessarily true, as the
placeholder may be reset outside _clearDragPlaceholder().
If this happens, the placeholder will temporarily stop working
for the "old" position (and permanently if it's the only position).
Just reset the position unconditionally.
https://bugzilla.gnome.org/show_bug.cgi?id=684618
Before, the text of those buttons were truncated when the text exceeded
the fixed width we had in the CSS.
Now, we give more horizontal space to the control buttons to match
the maximum text length of all buttons.
https://bugzilla.gnome.org/show_bug.cgi?id=696307
Now that we control our own destiny, I noticed that getResultsToDisplay
is the only user of this._notDisplayedResult, and it's called immediately
after setResults, which is the only thing that sets it. Just remove the
stateness entirely.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
Since the provider icon only appears in the list results, it makes
sense for that to be stored with the results class, rather than outside,
triggered by which sort of display it is.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
It turns out that this focus code broke sometime in the 3.6 cycle --
when updating results, the focus is always on the text entry, so this
never gets called. We'll eventually replace it with something that
keeps track of the focused result meta, but for now, remove it.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
While this is a very simple translation right now, soon enough it will
be so that it will have a less crazy "public" API and can do things like
cache result metas.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
Since the two paths that call this want to keep the actor in two different
states, it makes sense to just call the one function that's the same between
both individually.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
Right now, this doesn't give us very much, since IconGrid and StBoxLayout
have different APIs. But since we want to introduce result caching, it
makes to reduce the duplication we already have so we don't need to add
the code to do so in both places.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
pushResults, and the original async search API, were originally intended
so search results that weren't immediate could be added as they come in.
Since then, we've decided that the design of search results is that they
should finish at once with all results. Thus, the code was modified so
that pushResults always overwrote the current result set. As such, it makes
sense to rename the method so that the name matches the behavior.
https://bugzilla.gnome.org/show_bug.cgi?id=693836
Commit d0310bd745 blindly replaced global.overlay_group with
Main.layout.overviewGroup, but unlike the former, the latter is
hidden while the overview is not active, which makes it unsuitable
for the message tray's light box. In fact, with the removal of
global.overlay_group, there is no longer a container which may
be used both inside and outside the overview, so we can either
recreate the lightbox each time we show/hide the overview, or
use different lightboxes altogether; this opts for the latter.
https://bugzilla.gnome.org/show_bug.cgi?id=701097
This is a singleton object inside libibus which means that if we
destroy it (e.g. because ibus-daemon got restarted) then, other
library users, like the ibus gtk+ IM module that we also use
in-process, will break.
https://bugzilla.gnome.org/show_bug.cgi?id=699189
Currently we simply set the gsettings key when activating an input
source. This obviously introduces a time window, between the event that
activates the switch and when the switch is complete, under which key
events are being delivered to applications and interpreted according
to the previous input source.
The patches in bug 696996 introduce a DBus API in g-s-d that allows us
to know when an input source if effectively active. Using that and
freezing keyboard events in the X server until we hear back from g-s-d
we can ensure that events won't be misinterpreted after an input
source switch.
https://bugzilla.gnome.org/show_bug.cgi?id=697007
Passkey is the proper name to be used when we are pairing with a Bluetooth
2.1+ device. PIN are only used for older devices which is not the case
here, the message is only shown when dealing with 2.1+ devices.
https://bugzilla.gnome.org/show_bug.cgi?id=697661
Meta laters are invoked in reverse order of registration, so
having multiple laters propagating the geometry cause all but the
first one in the frame (which is usually wrong) to be ignored.
Instead, queue at most one later call, and use the last set geometry
in the callback.
https://bugzilla.gnome.org/show_bug.cgi?id=700853
In order to make sure that the struts and regions are initialized,
we need to make sure that we do with the conditions that nothing
is transformed, like the uiGroup, and that everything is visible.
These invariants broke during a fix to hide the UI until the
system background texture could be loaded. Now, reorder things so
that the system background is loaded, and then the UI is properly
loaded, and so on.
https://bugzilla.gnome.org/show_bug.cgi?id=696159
While the default style works well will a solid handle, using both
border and fill color would be desirable in classic mode. Add the
necessary (optional) style properties to allow this.
https://bugzilla.gnome.org/show_bug.cgi?id=697917
Instead of using the input mode, when the overview is not modal
it should use a Chrome-tracked actor, that is added to the input
region. Because the overview always takes pointer input when
visible, the actor is added at startup, and it is shown and hidden
as needed.
https://bugzilla.gnome.org/show_bug.cgi?id=700735
If we are passed an invalid file template, ShellRecorder.record()
will return a %NULL filename; as the Screencast DBus interface
expects a string return value, we cannot return the value unmodified
in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=700842
Variable names like "sourceNotificationStackDoneShowing" are too
long, and too undescriptive: this one points to a source, not a
notification stack that has been done showing.
As the close button of folder popups overlaps at the top, it ends
up being cut off if the folder is located at the very top of the
view. Fix this glitch by taking the button's overlap into account
in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=694371
We already take care of growing the view if open folders overlap
at the bottom, however folder popups may still end up being cut
off when opening above the source icon - if the popup is high enough,
its y coordinate will be negative and therefore outside the parent's
allocation. To fix, we can either make sure that folders pop up below
their source icon in that case, or adjust the parent grid's position
as necessary while a folder is open. This implements the latter.
https://bugzilla.gnome.org/show_bug.cgi?id=694371
If the view doesn't fill the available space, content should still
start at the top rather than the center - not least the positioning
code for folder popups assumes that, so set the appropriate expand
flags.
https://bugzilla.gnome.org/show_bug.cgi?id=694371
The main overview actor was made reactive to catch scroll-events
and propagate them; after some code shuffling, the actor that
catches scroll events ended up not being the same actor that's
supposed to propagate this, which broke using the scroll wheel
to switch workspaces.
https://bugzilla.gnome.org/show_bug.cgi?id=700595
Currently we show the workspace popup for relative targets ("up", "down"),
but not when targetting a specific workspace directly.
There is not really a good reason for that difference, and as we are about
to introduce a new shortcut to target the last workspace (which does vary
with dynamic workspaces), it makes sense to unify the behavior and always
show the switcher.
https://bugzilla.gnome.org/show_bug.cgi?id=659288
Currently we stop monitoring application usage when disabling the
'enable-app-monitoring' setting, but we still expose previously
gathered data in the app picker's frequent view. This is not what
users should expect, so hide the view in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=699714
Some keyboard spot a dedicated search key, which gnome-settings-daemon
currently handles by spawning gnome-search-tool. It makes a lot of
sense to promote the Shell's integrated search feature instead, so
expose an appropriate DBus method g-s-d can use.
https://bugzilla.gnome.org/show_bug.cgi?id=700536
This is technically a smidge slower due to the constant bisect insert,
but since this should only happen when we make a Wi-Fi dialog, it's
insignificant.
https://bugzilla.gnome.org/show_bug.cgi?id=700322
Commit 2499f2ed80 went back to using shell_app_activate() for
selecting an app, which favors windows on the current workspace;
this is the behavior we want for instance when activating a
launcher, but it's wrong for the alt-tab list - explicitly
request the first (e.g. MRU) window in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=700356
Since commit c84dc6254d, popup menus are closed automatically
when another menu opens (to catch the case where a menu is opened
by keyboard shortcut, which wasn't handled before). However in the
case of child menus, both child and parent are expected to be visible,
so handle this case explicitly.
https://bugzilla.gnome.org/show_bug.cgi?id=699678
Wrap new GtkMenuTracker API that adds an easy way to bind to
tracker items, and use it to add back support for submenus.
This also adds support for a submenu feature that we didn't
have support for before, action namespaces.
https://bugzilla.gnome.org/show_bug.cgi?id=700257
This pulls in new upstream API that Ryan will maintain, removing
code on our side.
Currently, our implementation of submenus will be gone, but this
will be fixed in a few commits.
https://bugzilla.gnome.org/show_bug.cgi?id=700257
When opening an application folder, it should take key focus to
allow for keynav; also, Escape closing both folder and app picker
is unexpected, it should only close the popup.
https://bugzilla.gnome.org/show_bug.cgi?id=695314
The point of a hash table is that you don't need to list all the
elements. To avoid that, keep a "clearableCount" in MessageTray,
which can be used by the message tray menu to show and hide the
clear item, and that is updated in constant time when sources
are added or removed.
https://bugzilla.gnome.org/show_bug.cgi?id=700194
Upon popMode, MessageTray will try readding all notifications
to their rightful parent, so we must tell NotificationBox to
relinquish them before st_bin_set_child() fails (leaving a dangling
child pointer and crashing at the next allocation)
https://bugzilla.gnome.org/show_bug.cgi?id=698812
Like screenshots, the screen recorder can be a useful tool in other cases
than being triggered by a keyboard shortcut. To account for that, export
a Screencast DBus API similar to the existing Screenshot interface.
https://bugzilla.gnome.org/show_bug.cgi?id=696247
Our built-in screen recorder is implemented as a component, so it will
just be disabled when the session mode doesn't allow screencasting.
However we will expose screencasting functionality on DBus as well, and
while it makes sense to restrict its availablity to the same modes as
the existing recorder, exporting/unexporting the service depending on
the session mode is not very consumer friendly.
For that reason, add an additional 'allowScreencast' property that for now
mirrors the availability of the 'recorder' component.
https://bugzilla.gnome.org/show_bug.cgi?id=696247
There is not always a clear distinction between code and style,
which is why the interface ends up being mostly unusable when we
end up without *any* style, for instance because the specified
application-stylesheet is corrupt.
Setting the default stylesheet in addition to the application-stylesheet
is no guarantee for non-default themes not messing up the interface, but
it should at least lower the risk ...
https://bugzilla.gnome.org/show_bug.cgi?id=700097
Metacity's Ctrl+Alt+Tab would include X11 windows
with hints like GDK_WINDOW_TYPE_HINT_DOCK and
GDK_WINDOW_TYPE_HINT_DESKTOP (there are more conditions, but that's a
good start). If we're in normal mode, those are visible and it's OK
to display those in the Ctrl+Alt+Tab order, but if we're in the lock
screen or the unlock dialog, they're not visible and it doesn't make
sense to focus them.
Bug: https://bugzilla.gnome.org/show_bug.cgi?id=699862
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Florian Müllner <fmuellner@gnome.org>
Some callers of the keyring prompt keep the dialog up while
processing the prompt. Allow the user to cancel the prompt
while in this state.
This is propagated to the caller, who can cancel the operation
in question when this occurs.
https://bugzilla.gnome.org/show_bug.cgi?id=682830
Only ACTIVE or ACTIVATING connections are important when deciding
what icon to show, don't fallback on any, possibly invalid or deactivating,
active connection object.
https://bugzilla.gnome.org/show_bug.cgi?id=676285
Commit e98eb57e3e added flags to expand the dialog's background
stack, which works fine with the current clutter-1.16 branch, but
breaks on clutter-1.14 (as shipped with GNOME 3.8).
Using an St.Widget with a Clutter.BinLayout fixes this, and is more
modern Clutter usage.
https://bugzilla.gnome.org/show_bug.cgi?id=699877
The popup currently has a fixed size based on monitor size. As a result,
the popup's content may overflow if its minimum size is larger than the
popup size. To prevent this, use min-width/min-height for the popup size
so that the popup can grow if necessary.
https://bugzilla.gnome.org/show_bug.cgi?id=696523
Currently a system modal dialog's actor hierarchy depends on whether
events should be blocked while the dialog is shown or not. Change
it to always contain a stack, to allow subclasses to add additional
background elements.
https://bugzilla.gnome.org/show_bug.cgi?id=694912
The date actors get destroyed and recreated on every date change which drops
key focus for the selected date. Restore key focus in such a case, but only
when the selected date was actually clicked. Whenever the next/prev month
buttons code is used (for scrolling, mouse click, or keyboard click), have
the corresponding button grab focus. Changing months currently causes the
calendar to update twice as the eventSource gets changed, so key focus gets
lost if it is on a date when the month changes.
https://bugzilla.gnome.org/show_bug.cgi?id=667434
The nautilus icon sticks out pretty badly among the symbolic
icons we use for other desktop components. This commit finds
windows of type DESKTOP, and uses the video-display-symbolic
icon for them.
https://bugzilla.gnome.org/show_bug.cgi?id=697914
Since we now have global.screen::in-fullscreen-changed, remove the
duplicate signal. To prevent ordering problems in connecting to
this signal, make inFullscreen a property-function of a new Monitor
object rather than a data property we tack on to a Rectangle object.
https://bugzilla.gnome.org/show_bug.cgi?id=649748
If we created a workspace after showing the view, we would never
set the geometry on it, which would cause an exception in the
window layout code and leave the DND state tracking in an undefined
state.
https://bugzilla.gnome.org/show_bug.cgi?id=699029
We put these "access point objects" in "this._networks" and
"this._activeNetwork", so let's rename it. This also makes
the fact that each "access point object" can contain multiple
access points a tiny bit less confusing.
https://bugzilla.gnome.org/show_bug.cgi?id=698918
NM is now a lot smarter about dealing with automatic connections, so just
create an empty connection and pass it to it. The only places where NM
requires connection settings is where we require explicit setup: Bluetooth
DUN, WPA-Enterprise and WWAN/VPN. These cases are already handled by
gnome-control-center, where complex configuration is handled, so remove
the automatic connection management for now and just let NM handle it.
https://bugzilla.gnome.org/show_bug.cgi?id=698918
We already do this when navigating between menus via mouse or
keynav, but miss cases where a menu is opened by other means,
for instance via a keyboard shortcut.
https://bugzilla.gnome.org/show_bug.cgi?id=686756
It makes sense to allow closing the app menu with the same shortcut
that is used to open it, so make it a toggle action and allow it
TOPBAR_POPUP mode.
https://bugzilla.gnome.org/show_bug.cgi?id=686756
Currently all keybindings are disabled while some popup menu is open.
However some keybindings may still be useful in some cases, so expose
GrabHelper's modal params parameter to allow specifying a keybinding
mode for particular menus.
https://bugzilla.gnome.org/show_bug.cgi?id=698938
Eval() is expected to return a boolean success value and a string result.
However when the function is disabled (via the development-tools setting),
we return null for the latter which is not a valid string value.
Return an empty string instead.
https://bugzilla.gnome.org/show_bug.cgi?id=698959
