- add some spacing between buttons
- move controls closer to the workspaces view (we'll need that space
for the message tray)
- fix the look of the scrollbar background
- adjust sizes of theme images
- some general CSS cleanup
https://bugzilla.gnome.org/show_bug.cgi?id=607872
This helps illustrate that expired notifications move to the summary view.
Animate the separate components of the message tray individually
as in http://www.gnome.org/~mccann/shell/mockups/20090630-demo
Based on the patch from Florian Müllner.
'workspace_relative' is used in workspace.js, not in workspacesView.js
Change a couple strings to have single quotes instead of double quotes to indicate that
they don't need to be translated.
When using the scroll wheel on the workspace indicators in single view,
make them use the same interpretation as the scrollbars: scrolling up
moves to the previous workspace, scrolling down to the next one.
https://bugzilla.gnome.org/show_bug.cgi?id=607824
Matching the 20091114 mockup, the default workspace view
is now a scrollable horizontal list, with a control to
switch between this and the previous grid view.
https://bugzilla.gnome.org/show_bug.cgi?id=593844
With the above preference set to true, the home and desktop entries
both represent the same location. Hide the desktop icon in that case
to avoid redundancy.
https://bugzilla.gnome.org/show_bug.cgi?id=606922
Added signal 'screen-size-changed' to ShellGlobal.
Connect to this signal in main.js and run the _relayout() method.
If Overview or calendar are visible when this signal emit, they will be hiding.
https://bugzilla.gnome.org/show_bug.cgi?id=584526
Make the event blocking done by Lightbox optional - leave it activated
when displaying the run dialog, but deactivate it when highlighting
windows in the overview.
https://bugzilla.gnome.org/show_bug.cgi?id=602774
The run dialog crashes when PATH contains non-existing directories (some
distros seem to do that ...).
Fix by filtering those out before setting up file monitors.
https://bugzilla.gnome.org/show_bug.cgi?id=606418
This bug happened because the thumbnail box was only maxed, not set to 256x256.
Standard window thumbnails would max out only the width to 256. Having a window with
its height greater than its width meant that the height would max out, but not the
width, causing thumbnails to look uneven.
https://bugzilla.gnome.org/show_bug.cgi?id=604963
Until now LookingGlass could only be closed using the Escape
key when the Evaluator class was active, as the relevant code
was associated to the text input field; this moves it into a
global event handler instead which works everywhere in LG.
We need to set the positions after the animation ends (and
fadeIn starts), not before when allocation is changed.
As the window may still be in motion when the overlay is
repositioned, it is safer to pass the calculated values
as parameters.
https://bugzilla.gnome.org/show_bug.cgi?id=602598
Consumer documentation will live at http://live.gnome.org/GnomeShell/Extensions
In terms of implementation; basically we load extensions from the well-known
directories. Add a GConf key to disable extensions by uuid. There is a new
option --create-extension for the gnome-shell script which takes a bit of
interactive input, sets up some sample files, and launches gedit.
No extensions UI in this patch; that will come later.
https://bugzilla.gnome.org/show_bug.cgi?id=599661
The high level goal is to separate the concern of searching for
things with display of those things; for example in newer mockups,
applications are displayed exactly the same as they look in the
AppWell.
Another goal was optimizing for speed; for example,
application search was pushed mostly down into C, and we avoid
lowercasing and normalizing every item over and over.
https://bugzilla.gnome.org/show_bug.cgi?id=603523
The distinction between the inactive and running was silly; just
have one class which can handle both running states. However for
a future search patch, we do want a separation between an icon which
just has icon + name + glow, and a well icon which does the menu
integration.
https://bugzilla.gnome.org/show_bug.cgi?id=603523
We has poor performance if we try to allocate hundreds of actors;
a bit more ideally, we'd keep pulling as much as we can, but really
no one should have more than 50 in the list as is now except on
unreasonably large screen sizes.
In the future we should change this to be pull-on-demand thing
in a chunked fashion.
https://bugzilla.gnome.org/show_bug.cgi?id=603522
Rather of calling .exists() synchronously for all documents, use
ShellDocSystem's async API to only stat docs we're showing.
Use the doc opening functionality in ShellDocSystem.
Also, more intelligently do redisplay(); don't recreate actors
for recent docs if we already have one for that URI. We may
need more intelligent caching here; if e.g. just the associated
application changes, we should probably reread the info.
https://bugzilla.gnome.org/show_bug.cgi?id=603522
Previously we had various things watching for notify::mapped so
we could be more lazy about updating non-visible actors, but it
was fairly ad-hoc.
The deferred work system unifies these callbacks, and also adds
a timeout so that we don't delay changes arbitrarily; this way we
avoid a storm of work if you stay out of the overview for a while,
then go in.
https://bugzilla.gnome.org/show_bug.cgi?id=603522
The altTab.js and app well code weren't sharing really
any functionality anymore; un-merge the appIcon code back
into appWell, and have a simple icon + text display for
altTab.
Port AppWell to St and CSS.
https://bugzilla.gnome.org/show_bug.cgi?id=602131
StClickable replaces ShellButtonBox. Reduce the number of
button-like things by deleting button.js.
To do so, add CSS style for the actitivies button.
https://bugzilla.gnome.org/show_bug.cgi?id=602131
It's nicer to have ShellDrawingArea as a St widget so it can
participate more cleanly in CSS styling, such as queuing a redraw
automatically on style changes, and allowing subclasses to use
CSS styling.
https://bugzilla.gnome.org/show_bug.cgi?id=602131
Some applications show a confirm dialog before closing, which the close
button happily ignores.
Detect newly created windows which are transient for the window we try to
close and switch to them.
https://bugzilla.gnome.org/show_bug.cgi?id=602532
We had previously been leaving the scaled workspace at a non-integral
position which didn't look obviously ugly, but now that we're
constraining the popup pane to the workspace size, we really need
to ensure that we're using integral positions here.
https://bugzilla.gnome.org/show_bug.cgi?id=601854
We need to check if the display actor is an instance of Shell.OverflowList
or St.BoxLayout to use the appropriate function for getting its child
with a given index.
Rather than the popup panes taking up the whole non-panel height,
constrain them to the height of the workspaces, which is also the
"dash content area".
https://bugzilla.gnome.org/show_bug.cgi?id=600734
Mostly a straightforward porting of style code to CSS, except
that various bits of other code referenced a few GenericDisplay
constants, so those needed to be ported as well.
Add some padding at the top between the close button and the items.
Center the text and description.
https://bugzilla.gnome.org/show_bug.cgi?id=600734
This should not be required, since glib correctly deals with such
applications by giving them the local gvfs path, and those
applications which do support URIs keep being able to use the URIs in
GFiles.
Based on original work by Guido Günther <agx@sigxcpu.org>.
https://bugzilla.gnome.org/show_bug.cgi?id=601287
It's very convenient to drill down through object properties and
be able to see exactly which portion of the screen those actors
correspond to, without trying to guess with the inspector tool.
Commit 94bd6f1718 introduced a trick
where we only do the heavy lifting for "redisplay" when we're mapped.
However, the search system wants to get the count of matched items,
and control the visibility of the display based on that. This introduces
a circularity; avoid it by forcing the search to do a redisplay.
In the future we should avoid this by separating out the "get matched
things for search" from "display list of things".
https://bugzilla.gnome.org/show_bug.cgi?id=600890
Places is one of the dash sections and it should be included in search results.
Factor out the code for getting and updating the information about places from
Places to PlacesManager.
Introduce PlaceInfo class that contains information about the place and can be
used by classes that display it in different ways. Rename classes so that their
names are consistent with corresponding classes in appDisplay.js and
docDisplay.js
https://bugzilla.gnome.org/show_bug.cgi?id=599125
This make it is easier for the user to figure out on which workspace the
windows are. For instance, terminals related to various activities and put on
different workspaces were previously displayed as an uniform list, with no
visible distinction between the ones from the current workspace and the others.
Now they are physically separated by a thin gray line.
This is also consistent with the way applications are displayed in the
AppSwitcher.
https://bugzilla.gnome.org/show_bug.cgi?id=597944
If we're not mapped, only queue up a redisplay. This avoids
e.g. changes in recent documents such as saving a file in GEdit
causing a lot of blocking I/O in the shell (we need to make
recent loading async as well).
https://bugzilla.gnome.org/show_bug.cgi?id=599560
First, fix a problem where though we intended to request a minimum
height of 0 for the docs content, we were actually requesting
spacing for all items.
On low resolution screens, we were still attempting to allocate
an item even when we were given 0 height.
https://bugzilla.gnome.org/show_bug.cgi?id=596984
For the purposes of determining which application is focused, don't
skip "uninteresting" windows. The old get_focused_window code
was used for usage tracking, but here we want reliable application
association.
Also convert a .text= to .set_text that was missed with the last
patch.
https://bugzilla.gnome.org/show_bug.cgi?id=599206
The two parts were mapping windows to applications, and
recording application usage statistics. The latter part
(now called ShellAppUsage) is much more naturally built on top of
the former (now called ShellWindowTracker).
ShellWindowTracker retains the startup-notification handling.
ShellWindowTracker also gains a focus-app property, which is
what most things in the shell UI are interested in (instead of
window focus).
ShellAppSystem moves to exporting ShellApp from more of its
public API, rather than ShellAppInfo. ShellAppSystem also
ensures that ShellApp instances are unique by holding
a hash on the ids.
ShellApp's private API is split off into a shell-app-private.h,
so shell-app.h can be included in shell-app-system.h.
Favorites handling is removed from ShellAppSystem, now inside
appFavorites.js.
Port all of the JavaScript for these changes.
https://bugzilla.gnome.org/show_bug.cgi?id=598646
Previously, we had ShellAppInfo, which contains fundamental
information about an application, and methods on ShellAppMonitor
to retrieve "live" information like the window list.
AppIcon ended up being used as the "App" class which was painful
for various reasons; among them that we need to handle window
list changes, and some consumers weren't ready for that.
Clean things up a bit by introducing a new ShellApp class in C,
which currently wraps a ShellAppInfo.
AppIcon then is more like the display actor for a ShellApp. Notably,
the ".windows" property moves out of it. The altTab code which
won't handle dynamic changes instead is changed to maintain a
cached version.
ShellAppMonitor gains some more methods related to ShellApp now.
In the future, we might consider changing ShellApp to be a GInterface,
which could be implemented by ShellDesktopFileApp, ShellWindowApp.
Then we could axe ShellAppInfo from the "public" API and it would
return to being an internal loss mitigation layer for GMenu.
https://bugzilla.gnome.org/show_bug.cgi?id=598227
If we had a pending search processing, finish it when the user
activates the entry.
This is a small conceptual change; the large diff is simply
moving the search implementation (unedited otherwise) from
an anonymous inline to a named function so it can be called
sanely in _activate.
https://bugzilla.gnome.org/show_bug.cgi?id=598111
POTFILES.in was missing some files.
"Activities" is not easily translatable into some languages, e.g. Russian,
so suggest using the word for "Overview" instead.
The down arrow is shown when an application with more than one
window is selected, but the window list is not always displayed.
This patch fixes the fact that the arrow was not hidden when one
focus an app with a single window when coming from an app with
multiple windows, if the window list was not displayed.
https://bugzilla.gnome.org/show_bug.cgi?id=597791
When we get a ClutterModifierType from Clutter, it might contain
bits not in the enumeration. See bug 59771 for a similar problem
with GdkModifierType.
Add a wrapper Shell.get_event_state() around clutter_event_get_state()
to mask these bits out and only return approved bits.
https://bugzilla.gnome.org/show_bug.cgi?id=597735
When tweening the overview in, tween it to appear on the correct
position of the primary monitor, not at (0,0); position
the 'backOver' actor to cover all monitors properly.
Reported by Rui Matos
https://bugzilla.gnome.org/show_bug.cgi?id=597721
Being able to display all applications in a category based on the search
string with a category name is generaly useful.
Prepare all the applications that match a search term based on their
category name up-front.
Remove unused this._appCategories and a call to non-existing
itemInfo.get_categories()
gdk_display_get_pointer() sometimes returns values for the mask that
aren't part of the GdkModifierType enumeration, which gjs doesn't like
(bug 597292). Work around that by adding a C wrapper that strips out
the extra flags.
https://bugzilla.gnome.org/show_bug.cgi?id=597559
Before we were badly leaking AppIcons by not disconnecting from the
window signal handlers when our actor got destroyed. This caused
us to repeatedly re-sort the windows for each AppIcon that
had ever been displayed with obvious bad consequences.
Besides simply chaining the signals to the lifetime of the AppIcon
actor, we also only do the sorting if we're mapped. This decreases
the amount of work to do in the not-mapped case.
https://bugzilla.gnome.org/show_bug.cgi?id=597120
Add a "size" parameter to allow changing the AppIcon size, and then
simplify the constructor by taking an object with parameters like
gobject-introspection constructors do, rather than taking a large
number of miscellaneous arguments.
https://bugzilla.gnome.org/show_bug.cgi?id=597498
Fix to not trigger a warning like:
JS ERROR: !!! WARNING: 'anonymous function does not always return a value'
by adding a return ""; at the end of the anonymous function.
https://bugzilla.gnome.org/show_bug.cgi?id=595661
For some unknown reason we were connecting to app-added and
app-removed on ShellAppMonitor in the AppDisplay class, which
never made any use of the data. Simply don't connect to those
signals for now. In the future we should have AppDisplay
be using the AppIcon class which will more correctly handle
dynamic changes.
In the AppWell, avoid doing the full relayout until we're
actually mapped.
https://bugzilla.gnome.org/show_bug.cgi?id=597169
Add String formatting by extending the String object with a
format method.
Now we can do stuff like "Text: %s, %d".format(somevar, 5)
This is required for proper translation of some strings.
https://bugzilla.gnome.org/show_bug.cgi?id=595661
This slightly changes the behaviour of the alt+tab window, this way:
when using alt-tab on a workspace that contains two or more windows from
the same window, the application selected when hitting alt+tab is the
currently selected application, but the highlighted window is the next
one.
Intended goal is to make it easier to cycle around windows of the same
application while not having to cycle through all the applications first.
https://bugzilla.gnome.org/show_bug.cgi?id=590563
By sorting the windows in the current workspace first, we ensure that
when using alt+tab to switch windows, we will pick a window on the
active workspace as the default focused window for the application if
the application has at least one window on the current workspace (that
is, if it is on the left of the app switcher separator).
This makes the behaviour of alt+tab more predictable for the user, as
an user will expect alt+tab to switch to the window he/she can see right
now rather than the one on the workspace he just left (presumably to do
something else on the workspace he's currently on).
https://bugzilla.gnome.org/show_bug.cgi?id=590563
Currently we use 0xefefefff as a background color for non ARGB tray icons,
which looks out of place (i.e does not fit the panel's background gradient).
Change it to 0x0b0b0bff to fix that.
https://bugzilla.gnome.org/show_bug.cgi?id=597148
We were actually showing a menu on button 3 before, but only
through a chain of coincidences. This patch explicitly supports
it and makes sure we show it immediately rather than after
a timeout. Pass the activating button in so that we only pop
down on that button.
https://bugzilla.gnome.org/show_bug.cgi?id=596371
Also have inactive applications pop up a menu.
Add/Remove from favorites is now in the menu.
Concatenate the favorites/not-favorites instead of having a gap only
if you happened to have a not-divisible-by-4 number of favorites.
https://bugzilla.gnome.org/show_bug.cgi?id=596371
This makes a visible distinction between the apps that only have minimized
windows on the current workspace and the ones that have no window on the
workspace.
https://bugzilla.gnome.org/show_bug.cgi?id=590563
This allows defining some custom policy for size allocation.
Currently, the minimum width is always used, but it can be tweaked
afterwards when a sizing policy has been defined.
https://bugzilla.gnome.org/show_bug.cgi?id=590563
We shouldn't show all sections each time the search is updated because that
breaks viewing search results for a single section and flickers headers for
sections with no results.
Fix displaying documents in the document browse and refreshing the selection
when the results have changed.
Make sure we are passing the appropriate flag to _redisplay() in GenericDisplay.
Make sure we set this._appsStale to true if there was a change in the
applications set.
Don't call _refreshCache() from the AppDisplay constructor.
Don't short-circuit the call to _refreshCache() from _redisplay() on initial load.
Rename _redisplayFull() to recreateDisplayItems() and remove adding an
actor to the actual result list in _addDisplayItem() because we redo adding the
actors to the list in _redisplayReordering() anyway to ensure that we add
them in the right order.
Based on a patch from Colin Walters.
Following the idea expressed in bug 590563 by mccann ("Minimized or
hidden applications should appear at the end of the list"), we should
also put applications that have no visible window in the active
workspace at the end of the alt-tab window list, after apps which have
minimized windows in the active workspace.
Miscellaneous fixes from review:
- Distribute calendar.js and the interactive test
- Make the pointless protection against leap seconds actually work
by starting in the middle of the day so that forward/back always
move a day.
- Use a variable instead of an inline '8' to know where to start
when removing old day actors.
- Remove a stray comment from the test
https://bugzilla.gnome.org/show_bug.cgi?id=596432
Make the calendar reactive and handle scroll events to change the month.
(GtkCalendar and hence the old gnome-panel calendar supported this and
it is apparently a handy way to flip through months.)
The padding is moved from the CalenderPopup to the Calendar so that the
scroll region extends all the way to the edge of the popup.
https://bugzilla.gnome.org/show_bug.cgi?id=596432
Previously, we initialized actor stacking order from the return
value of global.get_windows() once, which is defined to be in
stack order. However it was not updated later. Furthermore,
the way stackAbove was called from onAnimationComplete in
WindowClone was highly dubious, since there are lots of animations
which apply to the clones, and we want the stacking to be right
all of the time, not when some animation completes.
Fix this by connecting to 'restacked' on the screen and syncing
the clones.
I also snuck in another bugfix here; we weren't disconnecting
from the 'showing' signal handler, which had various bad
consequences.
https://bugzilla.gnome.org/show_bug.cgi?id=596263
js/ui/calendar.js: Generic calendar widget
tests/interactive/calendar.js: Basic test of the calendar
js/ui/panel.js: Add a pop-down from the clock that shows a
calendar widget. The pop-down is not menu-like to allow the user to
interact with an application while looking at the calendar.
gnome-shell.css: Add theming for calendar, calendar popup, and for
buttons on the panel
https://bugzilla.gnome.org/show_bug.cgi?id=596432
Remove the StTable specific methods to add actors:
st_table_add_actor()
st_table_add_actor_with_properties()
Since they shadow the generic ClutterContainer add_actor() method,
and patch in our add() convenience function as we do for
StBoxLayout.
https://bugzilla.gnome.org/show_bug.cgi?id=596811
Instead of using "Monospace", pick the users configured monospace font
name up from GConf. (This is a nice touch, but is more done here to
demonstrate that we can do it rather than for any great utility.)
https://bugzilla.gnome.org/show_bug.cgi?id=591245
* Style aspects like colors and fonts are moved into gnome-shell.css.
* Scrolling is adding using StScrollView.
Based on a patch from Colin Walters
https://bugzilla.gnome.org/show_bug.cgi?id=591245
ShellTheme replaces both StStyle and ccss_stylesheet_t.
The interface StStylable is replaced by usage of ShellThemeNode.
A concrete node class allows some significant optimizations of property
inheritance that would have been much more difficult to achieve with
the highly abstract pair of StStylable and ccss_node_t.
Some operations that were previously on StStylable (like the
::style-changed signal) are directly on NtkWidget.
Custom properties are no longer registered as param-specs; instead you
call directly into shell theme node to look up a length or color:
shell_theme_node_get_length (theme_node, "border-spacing", FALSE, &spacing);
The dependency on libccss is dropped, while preserving all existing
functionality and adding proper parsing and inheritance of font properties
and proper inheritance for the 'color' property.
Some more javascript tests for CSS functionality are added; workarounds for
a CSS bug where *.some-class was needed instead of .some-class are removed.
https://bugzilla.gnome.org/show_bug.cgi?id=595990
Install and distribute gnome-shell.css and theme images. They are moved
down from $datadir to $datadir/theme to avoid a weirdness where we have
images in $datadir and then also in $datadir/images.
(Also moved in the source tree to avoid adding another difference between
installed and uninstalled operation.)
https://bugzilla.gnome.org/show_bug.cgi?id=595989
js/ui/environment.js: Split out initial UI setup (Tweener initialization,
ClutterContainer monkey-patching) into a separate file we can import from tests.
tests/: Directory for various types of tests
tests/run-test.sh: Shell script that to run tests with an appropriate
environment set up.
tests/testcommon/: Common modules and data for tests
tests/interactive/: Interactive tests
tests/interactive/box-layout.js: A sample test of StLayout
https://bugzilla.gnome.org/show_bug.cgi?id=595987
Setting options for children added to StBoxLayout is not convenient
since we are missing the varargs methods of clutter_container.
Patch in:
child_set() - set properties of a child
add() - add a child and set properties (this is different from
clutter_container_add()! I think the deviation is
with avoiding the awkward name add_with_properties()
which is what might be expected. ClutterContainer
currently doesn't have a method like this at all.)
The code is written to allow patching into multiple ClutterContainer
classes but for now only StBoxLayout is patched, since it's the only
container we are using where we need to set options as properties.
https://bugzilla.gnome.org/show_bug.cgi?id=595419
Fix panel, app switcher, and looking glass to limit themselves to the
primary monitor, and run dialog to limit itself to the monitor
containing the currently-focused window.
The overview is also limited to the primary monitor now (with the
other monitors being blacked out), although the workspaces within the
overview are shaped like the full "screen" (the bounding box of all
monitors). To be fixed later.
https://bugzilla.gnome.org/show_bug.cgi?id=593060
Updated by Colin Walters <walters@verbum.org to monitor gconf
and use g_get_user_special_dir.
Signed-off-by: Adel Gadllah <adel.gadllah@gmail.com>
Signed-off-by: Colin Walters <walters@verbum.org>
https://bugzilla.gnome.org/show_bug.cgi?id=596933
Unlike icons in the application well, do not show the glow used to
indicate running apps. It is somewhat redundant here. These are all
running apps and it is fairly clear from the window list if there are
multiple instances available, according to mccann.
https://bugzilla.gnome.org/show_bug.cgi?id=590563
It slows things down, it's a slightly weird effect, and because the
window is still live while it's animating, you may see subwindows
being destroyed during the animation.
https://bugzilla.gnome.org/show_bug.cgi?id=596441
The text "Browse" used for the buttons on the dash section headers is a
little obscure and tends to translate into something that doesn't fit
well.
Going back to "More" will hopefully be more manageable in translation.
https://bugzilla.gnome.org/show_bug.cgi?id=596433
- Add spacing after Frequent, reduce it for the other app categories.
Put a small gap (one line) between Frequent and the other
categories to make it clear that it is something a little
different.
- Remove category icons from the applications menu.
Remove category icons; they aren't particularly helpful
(they are gone from the GNOME-2.28 menus too) and having them
in Applications Browse draws the eye to the wrong thing - the
category - rather than the right thing - the application icons.
https://bugzilla.gnome.org/show_bug.cgi?id=596435
The default AppIcon gray border color isn't very visible against a
dark gray background, but a white border looks too bold in the Well. So
allow the caller to override the AppIcon border color, so that the Well
can use gray-on-black and the AppSwitcher can use white-on-gray. (And
then revert the AppSwitcher back to the translucent gray background.)
https://bugzilla.gnome.org/show_bug.cgi?id=596337
The run dialog uses the "development_tools" gconf key to determine
whether funcions like "lg" or "restart" should be enabled, but it
failed to notice it on the run when said key changed. This commit
fixes this.
There are now 3 code paths in decreasing speed:
First, optimize subsearching more by just hiding the actors
that didn't match, since we know the ordering has to be right.
For initiating a search (or backspacing an existing one), again
instead of destroying and recreating actors, just temporarily
remove them and re-add them in the desired order.
Finally for when data has changed, use the old code path of
destroying all actors. (This itself could obviously be optimized
if we had a way to know that just one application changed, but
at the moment we don't).
https://bugzilla.gnome.org/show_bug.cgi?id=596119
We queue a 150ms timeout when the user starts typing to avoid searching
for the first keystroke. However, this caused us to change to the search
mode, but show the leftover state of the search displays from an
earlier search state.
Instead, just hide the results sections until we've actually performed
the current search once.
https://bugzilla.gnome.org/show_bug.cgi?id=596119
This is probably not the biggest optimization that needs to be
made at least for application searching, but we can optimize the
case where we're going from a search of "fi" to "fire" by just
re-searching the list of things we already had that matched "fi"
instead of looping over everything.
https://bugzilla.gnome.org/show_bug.cgi?id=596119
This fixes a regression where we weren't using the correct event
timestamps, because for both of these we were sending an XClientMessage
to ourself.
https://bugzilla.gnome.org/show_bug.cgi?id=596262
First eliminate the variable WELL_ITEM_HSPACING since it was 0
and thus effectively was not used.
Add a new variable WELL_ITEM_MIN_HSPACING which is the minimum
space between grid items we allow. When computing layout, allow
for a number of columns less than 4 by using the minimum item
size.
https://bugzilla.gnome.org/show_bug.cgi?id=595023
Based on a patch from Colin Walters.
