Colin Walters [Thu, 5 May 2011 19:21:58 +0000 (15:21 -0400)]
st-private: Fix memory leak
==13810== 11,360 bytes in 1 blocks are definitely lost in loss record 18,574 of 18,765
==13810== at 0x4005447: calloc (vg_replace_malloc.c:467)
==13810== by 0x5191882: standard_calloc (gmem.c:107)
==13810== by 0x51920A7: g_malloc0 (gmem.c:196)
==13810== by 0x4056201: blur_pixels (st-private.c:466)
==13810== by 0x40573B4: _st_create_shadow_cairo_pattern (st-private.c:710)
==13810== by 0x4070746: st_theme_node_paint (st-theme-node-drawing.c:856)
==13810== by 0x3FEFFFFF: ???
shell-app: Fix a case where last_user_time isn't updated.
When activating an uninteresting window, the last_user_time isn't updated,
because we aren't tracking the window that the user_time gets updated on.
Hack around this by setting the last_user_time in shell_app_activate when
activating an uninteresting window.
Dan Williams [Tue, 3 May 2011 17:21:45 +0000 (12:21 -0500)]
network: fix handling of AP flags and enhance for 802.1x
All WPA APs were getting set as WPA2 due to the check for privacy;
WPA/WPA2 APs *must* set the Privacy bit according to the standard,
so we'd never end up in the case for NMAccessPointSecurity.WPA.
Fix that, and also add flags for WPA[2] Enterprise which we'll
use a bit later for the first-time connect case for 802.1x enabled
access points.
Dan Williams [Mon, 25 Apr 2011 22:13:12 +0000 (17:13 -0500)]
network: simplify connection sorting by using libnm-glib functions
Instead of rolling our own code, use new libnm-glib functions to do
the same thing. Requires libnm-glib as of 779215c742bbe29a2c66202ec7e2e6d43edeb8ff (which will be part of 0.9).
Owen W. Taylor [Wed, 27 Apr 2011 11:50:10 +0000 (07:50 -0400)]
appDisplay: Fix off-by-one when incrementally adding application icons
A "cosmetic" code arrangement I requested in code review resulted
in one too few items being removed from the queue for each incremental
chunk of icons added. Fix.
We used used to indicate to the user the ability to move to another workspace
during dnd by highligthing the adjacent workspaces on hover.
This was done by changing the workspace's opacity to 200 and set it to
255 for the highlighted adjacent ones.
This is now no longer needed as the design was completely changed since
then (overview relayout; we no longer represent workspaces in the way
we did before) and introduces a bug where we don't properly reset the
opacity after the drag action, so just remove that code.
shell-app: Fix a case where last_user_time isn't updated.
When activating an uninteresting window, the last_user_time isn't updated,
because we aren't tracking the window that the user_time gets updated on.
Hack around this by setting the last_user_time in shell_app_activate when
activating an uninteresting window.
If the user typed "a", hit up, and pressed enter again, we wouldn't re-set the
history pointer to the end, so the broken navigation would instead go to the
entry before that.
This property represents that the widget is being labelled by an
actor. The name is label-actor to avoid problems with the current
StButton:label and StTooltip:label
WindowOverlay: Show close button after an animation
If a user is fast and mouses over a window while the workspace thumbnail
animations are playing, it can be frustrating when the close button won't
appear at the end of the animation.
Dan Winship [Mon, 18 Apr 2011 13:52:21 +0000 (09:52 -0400)]
altTab: fix Alt+Tab scrolling on initial display
The initial selection of the Alt+Tab dialog was happening before the
dialog was shown and allocated, and so the "do we need to scroll"
check used bogus coordinates. Fix by showing the dialog (and forcing
an allocation) first.
Adam Dingle [Thu, 21 Apr 2011 13:59:26 +0000 (06:59 -0700)]
Display dialogs on the primary monitor when no window has focus.
Before this change, we displayed dialogs on the monitor containing the
focused window if there was any, otherwise on monitor 0. We now use
the primary monitor rather than monitor 0 when no window has focus.
Previously, we skipped rebuilding device items in case the device
had already been seen, but this caused the connected switch not to
be updated. Now it has been refactored to update in case the device
changes, and to create only when the device is completely new.
PopupSliderMenuItem: intercept clicks outside the slider
Connect to button-press-event on the menu item actor, not on the
slider, so any part that is highlighted is also clicked. This means
that click on the left of the volume slider is a rapid way to mute.
NetworkMenu: keep wirelesss networks in predictable order
Adds a function that compares wireless networks and keeps them sorted
at all times. Order is: first already configured connections, then
first secure networks, then alphabtic. Also, the appearance of a new access
point no longer causes the whole menu to be rebuilt (but it still linear
searches for the position, I guess that could be skipped), which caused
the addition of more code for tracking the active access point.
Dan Winship [Wed, 6 Apr 2011 13:26:15 +0000 (09:26 -0400)]
panel: provide a consistent way to refer to panel items from lg
Rename a few Panel fields and add a _statusArea object pointing to the
status area PanelButton objects. You can now address each item
consistently from lg:
Florian Müllner [Tue, 29 Mar 2011 09:49:50 +0000 (11:49 +0200)]
app-menu: Update clip on icon size changes
To keep the app icon from overlapping the panel's (border-image)
border, a custom property for clipping the app menu icon's bottom
was introduced. But if the clip region is set before the initial
icon is set, the entire actor ends up clipped. Also due to the double
meaning of clutter_actor_get_height() (e.g. preferred height versus
allocated height), the clip region may end up too large and the icon
overlaps the panel's border-image.
Fix both problems by updating the clip region on size changes as
well, rather than on style changes only.
Colin Walters [Tue, 5 Apr 2011 16:48:44 +0000 (12:48 -0400)]
ShellWindowTracker: Ensure WM_CLASS remains canonical if it matches
I unintentionally made .desktop->pid association "win" over
WM_CLASS. Fixing this makes the case of ancillary .desktop file
entry points (e.g. gnome-control-center's various shortcut .desktop
files) correctly show System Settings, and not whatever the shortcut
is.
In the future I'd like to have a way to say "this .desktop file
is a shortcut, ignore me" or something.
Rui Matos [Wed, 23 Feb 2011 10:52:25 +0000 (10:52 +0000)]
altTab.js: remove Alt+Tab's special case
Alt+Tab's special case of "switch to the most-recently-used window even if
it's in the same app" is actually an hindrance for users to get a firm mental
model of Alt+Tab. Now that we have Alt+Above_Tab the special case is no longer
needed.
notifications: Fix order of title/banner for RTL text
Currently, the banner text is always located at the right of the
title - we need to swap those around for RTL locales. Rather than
using the locale directly to figure out the ordering, try to
determine the direction of the title label and let it control the
overall direction of the notification - this way, notifications
generally come out right when mixing scripts.
Fix search handling when typing multiple searches in the search box
In the current implementation of the search entry, when replacing a
selected search term with another term, the first character of the
replacement is prepended to the hint text rather than starting a new
search. This fix makes sure that the focus is not lost when changing
the selected search term.
When updating search results, the current result set is
recreated from scratch before setting the selection
highlight. This results in two style changes of the selected
item, and as a CSS transition is used to animate the style
change, the selected item flickers if it remains the same as
with the previous search term.
Fix by hiding the result set until the selection is set, to
avoid the transition in that case.
Dan Winship [Wed, 6 Apr 2011 16:54:47 +0000 (12:54 -0400)]
endSessionDialog: fix keyboard navigation
The addition of _backgroundStack to ModalDialog broke focus
navigation, because it was interposed between the focus group root and
all of the interesting content, but since it isn't an StWidget,
st_widget_navigate_focus() was unable to navigate through it. Fix this
by moving the focus root to this._dialogLayout (inside the
_backgroundStack) instead.
Additionally, in EndSessionDialog specifically, _initialKeyFocus
wasn't being set until after opening the dialog, so it was ignored.
Also change ModalDialog.close() to clear the _savedKeyFocus that
popModal() set, so that dialogs that are repeatedly closed and
reopened will have their focus revert back to _initialKeyFocus each
time.
Colin Walters [Wed, 30 Mar 2011 18:34:42 +0000 (14:34 -0400)]
dbus: Avoid losing org.freedesktop.Notifications on replacement
We weren't specifying _ALLOW_REPLACEMENT for anything except
org.gnome.Shell, which created a race - if the exiting process
didn't exit fast enough, the replacing process would fail
to get the name.
Colin Walters [Tue, 12 Apr 2011 21:29:05 +0000 (17:29 -0400)]
ShellWindowTracker: Follow transients for focus app
As a side effect of (see bug 642221), we no longer put docks or
transient windows into the hash table mapping windows to apps. The
"focused application" code relied on at least transients being in
there.
Fix this by calling the public API to map a window to an app, which
will at least follow transients. Whether we also want further
matching here (e.g. with window grouping) is another issue, but that
can happen as a different bug.
The icon set as secondary icon of the search entry depends on whether
a search is active or not - clicking the icon should reset the entry
only in the former case. This is implemented by connecting/disconnecting
the 'secondary-icon-clicked' signal when updating the icon, but an
additional signal connection was left-over when refactoring the search
entry, resulting in clicks on the inactive icon removing the hint text.
Fix by removing the stray signal connection.
Dan Winship [Tue, 5 Apr 2011 20:17:52 +0000 (16:17 -0400)]
st_label_set_text: no-op if the text is unchanged
If a caller sets an StLabel's text to what it already is (as, eg, the
clock menu does), do nothing. Unless the label is editable, in which
case, setting the text has a visible side effect (dropping the
selection), so we don't optimize that out.
Dan Winship [Fri, 1 Apr 2011 16:22:32 +0000 (12:22 -0400)]
lookingGlass: bring back the inspector icon
The change to StTextureCache for bug 644142 broke lg's inspector icon,
which was not specifying St.IconType.FULLCOLOR, but was relying on the
fact that SYMBOLIC (the default) would fall back to it. Fix the icon
by specifying FULLCOLOR explicitly.
(We should probably be using a symbolic icon here, but there is no
available icon with a select/pick/point-to/etc kind of meaning.)
Dan Winship [Mon, 4 Apr 2011 17:52:25 +0000 (13:52 -0400)]
src/Makefile.am: two fixes to the gnome-shell wrapper handling
Add an uninstall-hook to undo the effect of install-exec-hook, and add
an $(AM_V_GEN) to the rule that copies either gnome-shell-real or
gnome-shell-jhbuild to gnome-shell.
When more than one device exists, we need to reset the section title's
device to null, and in that case we must show nothing (neither the switch
nor the label, but an empty label is OK anyway). Also, we need to
update the device statusItem immediately when constructing it, as we
may not get any state-changed.
NMDeviceModem._createSection was not checking whether it should have
shown the connection list, resulting in status item shown even if
the device was in an invalid state.
Also, fix a logic error when creating the operatorItem and fix overriding
_clearSection protected method.
Dan Winship [Mon, 4 Apr 2011 12:31:55 +0000 (08:31 -0400)]
network: fix two warnings when removing a network device
NMApplet connects to each NMDevice's state-changed signal and stores
the signal handler id on the NMDevice itself. However, it was using
the same name as NMDevice itself was using to store the handler ID for
the underlying GObject's state-changed signal, thus overwriting it,
and resulting in *neither* signal handler getting removed if the
device went away. (This probably isn't a problem, since the device is
going away, but it causes a warning.)
Also, at least for WWAN devices, the device state changes to UNMANAGED
immediately before disappearing, but getStatusLabel() wasn't handling
that case and printed a warning instead. Fix that.
Owen W. Taylor [Sat, 2 Apr 2011 19:30:45 +0000 (15:30 -0400)]
network: deal with recycling of closed active connections
If you restart NetworkManager, then the list of active connections
is emptied, then comes back with the same GObjects in it. If the
_primaryDevice field isn't cleared on the object, then we won't
know we need to set it back on the device, resulting in the active
device not showing up in the menu.
NetworkMenu: fix updating the access point strength
A notify signal does not include the new value of the property in
its signature, so the handler was trying to compare a GParamSpec with
a number when updating. Fix it to always retrieve the value from the
object.
AppDisplay: refresh the view after refreshing sections
ViewByCategories._removeAll clears the sections, but the filter passed
to view still expects them to exists. Therefore, refresh the view
after the section has been rebuilt and the All filter reapplied.
Owen W. Taylor [Tue, 5 Apr 2011 14:54:28 +0000 (10:54 -0400)]
StScrollbar: clean up properly when unmapped during scroll
If we're unmapped (or destroyed) during a scroll, we want to clean
up the changes we've made to Clutter's event handling, remove our
signal handler, and emit ::scroll-stop.
projects / gnome-shell.git / log