The popup of the FolderView is now contained inside
the parent view, solving the overflow of apps with a ScrollView.
Also, solved a lot of bugs in popup/FolderView calculation
of position and size.
https://bugzilla.gnome.org/show_bug.cgi?id=706081
Add a property to also add the calculated spacing
around the grid.
This will allow FolderView to be aligned with the
main grid without cutting off any of the surrounding
boxPointer decorations or the close button
https://bugzilla.gnome.org/show_bug.cgi?id=706081
When we adapt the grid to different display sizes,
we don't want the number of displayed items to get
too small. In the future we will scale down icons to
make sure that the grid fits add least minRows
x minColumns items, but for now we only take the
properties into account when calculating the dynamic spacing.
https://bugzilla.gnome.org/show_bug.cgi?id=706081
Organize applications in AllView by pages using the new PaginatedIconGrid
added previously. Pagination is generally a better pattern for collections
than scrolling, as it better suits spacial memory.
Hook into AppDisplay's allocation function to communicate the available
size to the different views before child allocations - this is only
required by the paginated view (as pages must be computed before
calling get_preferred_height/get_preferred_width), but doing it for
all views will guarantee that their dynamic spacing calculation is
based on the same values.
https://bugzilla.gnome.org/show_bug.cgi?id=706081
Since the parameter of the function is the width, reflect that in
the function name. Also, since we are counting columns, not only
children for each row, reflect that in the function name also.
https://bugzilla.gnome.org/show_bug.cgi?id=706081
The new PaginatedIconGrid class acts as a container for pages.
So the new class provides the container behaviour and some
useful functions like positions of pages, number of pages, etc.
But, it doesn't add indicators of the pages and doesn't manage
the scroll of the pages, neither any management of the pages
like in which page currently it is, etc.
https://bugzilla.gnome.org/show_bug.cgi?id=706081
This method, which accepts a .desktop filename, is used to highlight
a specific application in the overview, for example because it has
just been created or installed.
https://bugzilla.gnome.org/show_bug.cgi?id=654086
We added special code to sort each row in the overview so that
windows were less likely to cross lines, but the awkward control
flow meant that everything but the last row got sorted.
https://bugzilla.gnome.org/show_bug.cgi?id=707197
Activating the overview is fairly easy (hot corner, <super>), so doing it
automatically after closing the last window on a workspace does not save
a lot of effort; it does result in a surprising context switch when the
user does not expect the behavior.
https://bugzilla.gnome.org/show_bug.cgi?id=662581
We watch changes in the VPN state, not the active connection state,
so if we use the active connection state, we might miss an update
(because the VPN property is notified before the other one)
https://bugzilla.gnome.org/show_bug.cgi?id=706262
Descriptions are only added after all devices are read (thanks
to the disambiguation in libnm-gtk), but we use them immediately
when we call _sync() in various points (such as checkConnection())
https://bugzilla.gnome.org/show_bug.cgi?id=706262
Right now we only show the session menu button when verifying,
but we should also show it when verification is failed or we
can end up in situation where the session menu disappears during
an authentication retry.
https://bugzilla.gnome.org/show_bug.cgi?id=707064
This commit consolidates the styles of the various
message types into one 'login-dialog-message' style
and then adds additional styles on top to cover the
differences.
This allows us to give the message label an initial
style so that is padded properly before any messages
are displayed.
https://bugzilla.gnome.org/show_bug.cgi?id=706670
Build gnome-shell for x11, and gnome-shell-wayland for wayland
(as well as the associated libgnome-shell and libgnome-shell-wayland).
The first one links to libmutter, the second to libmutter-wayland.
libgnome-shell and libgnome-shell-wayland are now compiled from
libgnome-shell-base (with all sources that are independent of mutter),
libgnome-shell-menu (with the copy-pasted gtk sources), plus the
sources that use mutter API
https://bugzilla.gnome.org/show_bug.cgi?id=705497
These cause annoying allocation cycle warnings, and it's simpler to
just express our desired layout in terms of nested containers.
Adapt the theme to match as well.
https://bugzilla.gnome.org/show_bug.cgi?id=706843
We slide the shield over it, so the animation is rarely seen, and
since no other actor is under the lock screen, the not-cleared stage
can show through, causing weird issues when trying to blend.
https://bugzilla.gnome.org/show_bug.cgi?id=706841
When we implemented the new designs, we lost the ability to suspend
from the system menu. Re-enable this ability by re-adding the hidden
"Alt" shortcut item.
https://bugzilla.gnome.org/show_bug.cgi?id=706612
Right now, we rely on PAM to ask for the username if disable-user-list
is TRUE. This is suboptimal because it means we can't check if we
should show a session menu.
This commit changes disable-user-list==TRUE to ask for a username up
front, rather than have PAM do it.
https://bugzilla.gnome.org/show_bug.cgi?id=706607
We show a lightbox when we suspend, to animate the fading to black
caused by turning off the monitors, but we need to hide it when
coming back, otherwise the user is just staring at a black screen
it until he moves the mouse or presses a key.
https://bugzilla.gnome.org/show_bug.cgi?id=706654
Sometimes gnome-session hands us a bad object path for JIT inhibitors
it creates for XSMP clients. While this is a bug in gnome-session, we
shouldn't show an empty-looking dialog here.
https://bugzilla.gnome.org/show_bug.cgi?id=706612
These don't go through gnome-session, so they don't properly update
its state machine. We should use these in the future when we want to
use logind user sessions, but for now, they're just a trap.
https://bugzilla.gnome.org/show_bug.cgi?id=706612
A D-Bus service can export more supported interfaces than the
shell cares about. In those cases, we avoid creating proxies,
but neglect to finish things up so the object manager class
knows it can mark itself loaded.
This commit makes sure we do the proper finishing, so the object
manager still loads in the face of unsupported interfaces.
https://bugzilla.gnome.org/show_bug.cgi?id=706542
When locking manually (or locking with an animation), fade the
screen to black after a small timeout. This provides a smoother
experience, instead of abruptly turning off the screen.
https://bugzilla.gnome.org/show_bug.cgi?id=699112
Unfortunately, display configuration can and does fail, due
to unspecified HW constraints, drivers bugs, unsupported exotic
configurations or just bad luck.
So when the user makes a change in the control center, show
a dialog asking him if it looks OK, and revert back after 20 seconds
otherwise.
https://bugzilla.gnome.org/show_bug.cgi?id=706208
All our modal dialogs are given a fixed width and grow vertically
as necessary. Set the request mode accordingly, so that wrapped
labels are considered correctly during size request, and not only
at allocation time (where they'll either take away from the padding
or even cause the dialog to overflow).
https://bugzilla.gnome.org/show_bug.cgi?id=704015
The only point of using a custom container here was to prevent StBoxLayout
from enforcing the wrong request mode based on the orientation. With that
issue fixed, we can simplify the checkbox widget significantly.
https://bugzilla.gnome.org/show_bug.cgi?id=703811
Replace more direct XFixes usage with a the appropriate abstraction
API from mutter, which is guaranteed to work in wayland too.
It doesn't yet replace pointer position tracking, although probably
it should.
Also, because now we're using Mutter API, we lose the standalone
test case.
https://bugzilla.gnome.org/show_bug.cgi?id=705911
Mutter now includes an object with the same purpose and functionality
as ShellXFixesCursor, so we can replace our XFixes code with it
and work under wayland too.
https://bugzilla.gnome.org/show_bug.cgi?id=705911
Triangles should be flipped in RTL. This is the easiest way to do it that
doesn't rely on modifying the rotating logic, though it is a bit hacky since
the ClutterActor "scale-x" property technically considers the lower bound
to be 0. It works, though.
GrabHelpers use a 'captured-event' to steal events and emulate
modality or grab-like semantics. There can be issues when you try to
use multiple GrabHelpers stacked on each other. As Clutter follows
the DOM-like semantics of "first come, first serve", when a second
GrabHelper connects to 'captured-event', its callback will only be
processed *after* the first GrabHelper's callback is called.
This breaks the expectation of narrowing modality where new modals
take priority over the old ones.
Solving this globally in a cleaner manner would require a rewrite of
pushModal/GrabHelper. As a stopgap fix for now, use one shared
'captured-event' handler between all GrabHelper instances, and
delegate to the individual GrabHelpers.
https://bugzilla.gnome.org/show_bug.cgi?id=699272
This commit detects when a user inserts a smartcard,
and then initiates user verification using the gdm-smartcard
PAM service.
Likewise, if a user removes their smartcard, password verification
(or the user list depending on auth mode and configuration) are initiated
https://bugzilla.gnome.org/show_bug.cgi?id=683437
gnome-settings-daemon monitors smartcard insertion and removal
events on the system and then exports a model of the current
smartcard topology over the bus using the D-Bus ObjectManager interface.
This commit adds the support code needed in gnome-shell to talk to
the gnome-settings-daemon service.
A future commit will use this code to inform the login screen
when a user inserts a smartcard (so it can react appropriately)
https://bugzilla.gnome.org/show_bug.cgi?id=683437
The D-Bus ObjectManager interface is fairly recent addition to the
D-Bus specification. Its purpose is to provide a standardized way
to track objects dynamically coming and going for a service, and
to track capabilities dynamically coming and going for those objects
(by means of interfaces).
This commit adds the requisite code needed to make use of the
ObjectManager interface.
It will ultimately be needed to implement smartcard support in the
login screen.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
The duplication makes the function look a lot more complicated
than it actually is.
This commit moves the common code to a new _startService function.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
Some pam modules prompt without expecting the user to type
an answer back (e.g. "Please swipe finger"). We need to
emit prompted in this case too, so the the dialog will get shown.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
Currently, fingerprint authentication is always a secondary thing.
If a user wants to swipe their finger when the computer is asking
for a password, so be it.
This commit paves the way for making fingerprint auth optionally
be the main way to authenticate. Currently there's no way to enable
this, but in a future commit will honor
enable-password-authentication=false
in gsettings.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
Right now, the primary way a user logs in is with
a password. They can also swipe their finger, if their
fingerprint is enrolled, but it's expected the fingerprint
auth service won't ask questions the user has to respond to
by typing. As such, we ignore questions that comes from
anything but the main auth service: gdm-password.
In the future, if a user inserts a smartcard, we'll want
to treat the gdm-smartcard service as the main auth service,
and let any questions from it get to the user.
This commit tries to prepare for that eventuality by storing
the name of the default auth service away in a _defaultService variable
before verification has begun, and then later checking incoming
queries against that service instead of checking against
string 'gdm-password' directly.
Of course, right now, _defaultService is always gdm-password.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
This commit introduces a new BeginRequestType enum which gets
passed to the 'reset' signal to specify whether
a username should be provided to the begin() method and changes
the loginDialog to comply.
Currently, the signal only ever gets emitted with
AuthPrompt.BeginRequestType.PROVIDE_USERNAME
but that will change in the future when providing smartcard
support.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
We currently emit "failed" any time the UserVerifier is reset,
and user verification didn't succeed prior.
A more conceptually clear time to emit "failed" would be if
the UserVerifier is reset and user verification failed prior,
and to emit "failed" if the user cancels unlock.
This commit restructures things to do that. Aside from being
more conceptually clear, it also lays the groundwork for us
to be able to reset the unlock screen without failing.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
authPrompt.reset() currently only leaves the authPrompt in a
sane state if the user isn't verifying.
This commit makes sure to cancel verification if a reset happens
while verification is in process.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
If we don't have a connection at startup or we transition from
having a connection to not having a connection, we need to make
sure we hide the correct indicators.
_updateState has a lot of variables that sort of gunk up the
code and make it more unreadable than need be. Clean up the logic
a lot by moving those variables into the places that they actually
matter, renaming them to remove prefixes, and remove some conditions
that are always met.
Right now the code chooses to animate based on whether or not the
notification was "removed", which is quite a sketchy subject. For
now, add an additional case so that we don't animate when we transition
to the lock screen.
When the triangle rotates (when sub-menu is expanded), it seems as if
the triangle pivots from one corner even though rotation center is set
to Clutter.Gravity.CENTER. Hence the rotation center is set nearer to
the edge than to the corner ([0.3, 0.5] instead of [0.5, 0.5]) so that
it doesn't appear odd.
Also pivot_point is used instead of rotation_center_z_gravity as it is
deprecated.
http://bugzilla.gnome.org/show_bug.cgi?id=703109
This will replace the indicator painted on the stage right now.
This unfortunately does not work for the recorder triggered by the
keybinding -- we'll simply replace the in-shell code with a keybinding
powered by gnome-settings-daemon.
The existing app menu was a kludge of legacy code that tried to manage
a bunch of state, and had a number of issues:
* It didn't properly manage visibility when combined with multiple
apps and the overview.
* It didn't properly manage reactivity when tabbing away from a busy
app to another app.
* It didn't properly disconnect signals when going from one app
to nothing.
and countless others. Rewrite it to use the new "sync" code pattern,
where we centralize all state management and do transitions from that,
rather than strange and quirky control flow.
https://bugzilla.gnome.org/show_bug.cgi?id=705898
Make the lock dialog group reactive, to intercept any events
before they go to the actors below.
In the future, we may restructure our chrome to have a clear
layer system, but for now it fixes a security issue in the lock
screen (you can see the contents of the windows by dragging
if the screen was locked with the overview active)
https://bugzilla.gnome.org/show_bug.cgi?id=705840
There's only two uses of the parameter left, which can easily be added as a
separate line below. Since it's really a private interface meant for the
indicators, make it private as well so external users are less likely to
use it.
https://bugzilla.gnome.org/show_bug.cgi?id=705845
To align the arrows, we need to allocate panel buttons the full
height of the tray. Fix up all of the panel buttons to support this,
and align the arrows in the middle.
https://bugzilla.gnome.org/show_bug.cgi?id=705845
Swap out the implementation of SystemIndicator with a dummy,
and build the aggregate menu. At the same time, remove the
poweroff and login screen menus, as those were fake aggregate
menus beforehand.
We lose some flexibility as we lose session-mode-based menu
layout, but as each component of the aggregate menu is supposed
to be "smart" in response to updating itself when session
state changes, I believe it's better than a declarative model.
https://bugzilla.gnome.org/show_bug.cgi?id=705845
We can't silently replace the old behavior of separate status
icons into a new system. Replace SystemStatusButton with a new
SystemIndicator class which will allow for the flexibility we
need. For now, make it a subclass of Button so that it mostly
feels the same, but we'll soon be swapping it out with a dummy
implementation that the aggregate menu will use.
I think the code cleanup here is worth it.
https://bugzilla.gnome.org/show_bug.cgi?id=705845
This code is too complicated to keep, and the last straw came after the
fixed width menu in the aggregate menu design.
This will break some existing popup menus that rely on the fixed width,
but this will soon be replaced with the aggregate menu. We'll also soon
clean this up further by replacing PopupBaseMenuItem's custom layout code
with an StBoxLayout.
https://bugzilla.gnome.org/show_bug.cgi?id=705845
onAskQuestion has this code:
if (this.verifyingUser)
this.cancelButton.show();
else
this.cancelButton.hide();
but onAskQuestion can only be called when this.verifyingUser is true.
Also, cancelButton is public, and it only ever otherwise gets hidden
from callers.
This commit drops mucking with cancelButton visibility, leaving it
entirely up to the callers to deal with.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
Showing the new message at full size marks an abrubt change and looks
bad. Instead, gradually animate from 0px to full natural height.
Includes hacks to workaround flickering scrollbars while the animation
is in progress.
https://bugzilla.gnome.org/show_bug.cgi?id=687660
If that fails (which only ever happens in initial-setup mode, which
has no unlock or login dialog), we don't want to go ahead with
whatever we were doing.
https://bugzilla.gnome.org/show_bug.cgi?id=701848
If we don't remove the animation, we might leave a pending call
to _lockScreenShown() which would confuse our state tracking into
thinking we're active when we're not.
https://bugzilla.gnome.org/show_bug.cgi?id=700901
Using a signal handlers causes us to depend on connection order, but
we need the message tray code to run last, so it can notice that
notifications are destroyed when hiding the boxpointer and skip
the broken animation.
https://bugzilla.gnome.org/show_bug.cgi?id=686855
If the first question asked to a user is from the
shell and not from the PAM service (i.e. Username: ),
then we'll save what the user types until PAM asks
a question and then try to send it to PAM.
This commit makes sure the preemptive answer can be used
before the PAM conversation gets started, and makes sure
to discard the preemptive answer if we're not expecting it.
https://bugzilla.gnome.org/show_bug.cgi?id=705370
Right now we have two booleans that specify when user verification
is happening and when it succeeded, respectively.
This commit consolidates them into one AuthPromptStatus enumeration.
This clean up will allow us to check for verification failure more
easily.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
The only time we ever call _reset directly is when
detecting changes to disable-user-list. We can implicitly
trigger a reset for this case, just as easily by calling
this._authPrompt.reset()
This commit makes that change for consistency and to make
it easier to adjust the authprompt workflow later.
https://bugzilla.gnome.org/show_bug.cgi?id=683437
When we shift workspaces to create a blank one for a window or
application, all of the window actors are shifted down. However, some
of these window actors are transient windows attached to a main window.
When these windows are moved to a different workspace, the main window
is moved along with it. When the main window is moved, these windows
are also moved. This creates a double move of the windows.
This double movement leads to unexpected results where workspaces are
collapsed and windows are in incorrect positions.
This patch prevents movement of these transient windows, only grabbing
the main (ancestor) windows to move to a different workspace.
https://bugzilla.gnome.org/show_bug.cgi?id=705174
Right now the whole authPrompt spreads out if a PAM message
comes in that longer than the entry.
This commit changes it to wrap instead, by forcing the
auth prompt to be a fixed width (slightly bigger than
the entry width was sized to previously).
https://bugzilla.gnome.org/show_bug.cgi?id=705037
Remove the Wi-Fi chooser from the menu and put it in a dialog instead.
This frees up the submenu to simply have three items: an rfkill toggle,
a button to show the dialog, and a button to show network settings.
Ideally, we'd autodetect the "needs network" case by user initiation
and automatically show the dialog if needed, but lower-level plumbing
is neccessary, so the menu item to show the dialog is an acceptable
compromise instead.
This is a part of the new system status design, see
https://wiki.gnome.org/GnomeShell/Design/Guidelines/SystemStatus/
for design details.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
Since the network section of the aggregate menu will be shown in the lock
screen, we need to ensure that users can't tweak with network settings or
anything like that.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
Replace NMNetworkMenuItem with NMConnectionItem, based on
NMVPNConnectionItem, and replace NMDevice with NMConnectionSection
and NMConnectionDevice.
Since this rips apart NMDevice, and since wi-fi should not be
connection-based, we'll temporarily remove NMDeviceWireless. We'll
add it back in a later commit, along with the new Wi-Fi dialog.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
Instead, just add them after they're constructed. This allows us to
not have to pass the connections to each device, and prevents issues
with having to enumerate the connections in the middle of construction.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
This is a part of the new system status design, see
https://wiki.gnome.org/GnomeShell/Design/Guidelines/SystemStatus/
for design details.
Note that this does have an interesting side effect of not showing
network connectivity status on wired. This is intentional, and error
states will still be shown in the top bar when they happen.
This also means that if you're connected to both wired and wireless,
even though wired is the default route, we'll first notice the wireless
active connection, and we'll show that in the top bar. New NM API that
will help figuring out the active connection of the default device is
being implemented to stop this from happening.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
The code is complicated by requiring overflow, and in order to incrementally
improve the code to match the designs, remove overflow.
In the new design, we'll have a fixed number of menu items, and Wi-Fi
will be done by a separate design, so we can't be too concerned with
the menu not fitting on the screen.
This is a part of the new system status design, see
https://wiki.gnome.org/GnomeShell/Design/Guidelines/SystemStatus/
for design details.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
According to Dan Williams, if firmware is installed the device
will disappear and reappear, and this is unlikely to change any
time soon. Just make our lives easier by removing the tracking.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
I intended to make a few code cleanups, but I apparently forgot
to hook up _updateAccessPoint. Merge it with _activeApChanged,
which is where the notify::active-access-point signal is actually
hooked up to.
https://bugzilla.gnome.org/show_bug.cgi?id=704670
As we only reload search providers on startup or when the sort order changes,
and given the small number of search providers we'll actually load, I doubt
we'll see any speed decrease.
The simplicity of synchronous code is also much clearer, and fully avoids
all the possible bugs about in-flight requests or similar.
This also prevents issues with multiple search providers showing up at once,
which happen when multiple requests to reload search providers get called
immediately, with the existing in-flight async requests never cancelled.
https://bugzilla.gnome.org/show_bug.cgi?id=700283
When we reload the remote search providers, we currently try to remove
all remote providers, and then re-scan. It turns out that we sometimes
remove the wrong providers from the remote provider list, causing us to
have some providers not correctly unloaded.
https://bugzilla.gnome.org/show_bug.cgi?id=700283
When a ShellUserVerifier is asked to verify a user at the login
screen it will transparently first try to reauthenticate the user
against an existing session and then fall back to logging a user
into a new session. The former is used for user switching.
It's useful to know which type of verification is happening, so
the next button can be made to say "Unlock" instead of "Sign In" when
a user is already signed in.
This commit exports a new "reauthenticating" property on the
ShellUserVerifier that the auth prompt checks when deciding which
label to use for its next button.
https://bugzilla.gnome.org/show_bug.cgi?id=704795
If there are no messages in the queue and a user starts to
type then we can safely hide the message label since the
user has probably already read it.
This fixes a weirdness where "Incorrect Password" messages stay
around, even as the user types in the new correct password.
https://bugzilla.gnome.org/show_bug.cgi?id=704817
Similar to our ClutterContainer monkey-patching, we can add some
convenience to existing ClutterLayoutManagers:
- hookup_style() to bind layoutManager properties to CSS properties
- child_set() to set child properties
https://bugzilla.gnome.org/show_bug.cgi?id=703905
Jasper removed the ShellGlobal:stage-input-mode property after its
"last" use was removed. Adapt the (hopefully) really last use of the
property to the recent input changes.
https://bugzilla.gnome.org/show_bug.cgi?id=704095
There's quite a bit of duplicated code between the login dialog
and the unlock dialog dealing with the various signals from the
ShellUserVerifier.
This commit moves that duplicated code into the AuthPrompt.
https://bugzilla.gnome.org/show_bug.cgi?id=704707
The point of fading the icon is to make the text displayed over the
icon more legible. In RTL layouts, the text is displayed on the left
of the icon, so fading the right-hand-side of the icon doesn't work
well.
https://bugzilla.gnome.org/show_bug.cgi?id=704583
This is a regression from splitting the slider out that never got fixed.
Restore the previous (useful) behavior by adding a public API to the
slider that lets us pass an event through.
https://bugzilla.gnome.org/show_bug.cgi?id=704368
We've long had the hasWorkspaces property, but it doesn't seem like
it was ever used. Implement it so that we don't have workspaces in
initial-setup mode.
Since it's difficult to make it change at runtime with a decent set
of semantics, and we never expect that to happen, don't bother
implementing it dynamically.
https://bugzilla.gnome.org/show_bug.cgi?id=698593
commit ea02380c15 made the login
screen stop using ModalDialog. It makes sense for the unlock
code to also stop using ModalDialog, too (for similar reasons).
Now that the login screen's auth prompt code has been separated
out, the unlock dialog can use it to get the buttons and spinners
etc, that it was previously getting from ModalDialog.
This commit drops the ModalDialog usage in the unlock dialog, and
makes the unlock dialog use GdmUtil.AuthPrompt instead.
https://bugzilla.gnome.org/show_bug.cgi?id=702308
Right now there is a lot of duplicated code between the unlock
dialog and the login dialog.
This commit moves the login dialog's auth prompt to a separate
class, so that it can (in a subsequent commit) be used by the
unlock dialog.
https://bugzilla.gnome.org/show_bug.cgi?id=702308
A bug got introduced when moving the login dialog away from modal
dialog, such that it listens for escape key presses in a mouse
event handler instead of a keyboard event handler.
This commit fixes that code to correctly listen for key-press-event
instead of button-press-event.
https://bugzilla.gnome.org/show_bug.cgi?id=702308
