We explicitly include NoDisplay applications in the ShellAppSystem because
we want app tracking for them, but we explicitly filter NoDisplay applications
out when showing them to the user because we don't want to show them to the
user. We also based our "All" apps view on a flattened list of apps. While
we did check for NoDisplay on the app item itself, we didn't check against
its parents. Refactor the app display view to not use a separate flat list
of applications, but instead a concatenation of all the applications in all
the loaded categories.
https://bugzilla.gnome.org/show_bug.cgi?id=658176
The function may return FALSE without setting the GError, so don't
assume it is set to prevent a crash in that case. While at it, free
the GError we were leaking before.
https://bugzilla.gnome.org/show_bug.cgi?id=670418
It seems that Debian has their own prefixes in something like
debian-xterm.desktop. To properly do application matching in these cases,
we need to strip the debian- prefix.
https://bugzilla.gnome.org/show_bug.cgi?id=665647
We originally OR'ed search terms and favored results which matched
multiple times to get more relevant results. When changing search
to AND search terms, the semantics of "multiple matches" were
changed to refer to a single term matching multiple criteria (name,
executable), which seemed like a good idea at the time.
However in practice this just results in applications whose
user-visible name matches the executable name on disk being
favored over applications using a more generic name, which
isn't too useful (in particular when taking usage frequency
into account).
https://bugzilla.gnome.org/show_bug.cgi?id=623372
Application search results are internally categorized in four sets,
multiple and single prefix matches and multiple and single substring
matches. Each set is currently sorted alphabetically by application
name when concatenating the sets to the final result.
Change the last step to sort each set by usage frequency instead,
which is more likely to favor the most relevant match than
"arbitrary" alphabetic order.
https://bugzilla.gnome.org/show_bug.cgi?id=623372
Commit 0af108211c introduced a
regression where applications that appear in multiple categories were
duplicated in the "All Apps" list, because we switched from
uniquifying on desktop file ID to the GMenuTreeEntry.
Switch back to keeping the set of apps based on ID. To flesh this
out, we keep the ShellApp instance for a given ID around forever, and
when we're loading new contents, we replace the GMenuTreeEntry inside
the app. That means callers still get new data.
We still keep around the running app list, though we could just
recompute it from the app list now.
https://bugzilla.gnome.org/show_bug.cgi?id=659351
The apps and settings loading code duplicated the part to traverse a
GMenuTree. Unify this by adding a new function to return a flattened
set.
This will also be useful for a future change to how we store apps -
this way we can look at both the current set of apps and the new set.
https://bugzilla.gnome.org/show_bug.cgi?id=659351
This patch fixes the "apps vanish from alt-TAB bug".
If a "package system" rips away and possibly replaces .desktop files
at some random time, we have historically used inotify to detect this
and reread state (in a racy way, but...). In GNOME 2, this was
generally not too problematic because the menu widget was totally
separate from the list of windows - and the data they operate on was
disjoint as well.
In GNOME 3 we unify these, and this creates architectural problems
because the windows are tied to the app.
What this patch tries to do is, when rereading the application state,
if we have a running application, we keep that app around instead of
making a new instance. This ensures we preserve any state such as the
set of open windows.
This requires moving the running state into ShellAppSystem. Adjust
callers as necessary, and while we're at it drop the unused "contexts"
stuff.
This is just a somewhat quick band-aid; a REAL fix would require us
having low-level control over application installation. As long as
we're on top of random broken tar+wget wrappers, it will be gross.
A slight future improvement to this patch would add an explicit
"merge" between the old and new data. I think probably we always keep
around the ShellApp corresponding to a given ID, but replace its
GMenuTreeEntry.
https://bugzilla.gnome.org/show_bug.cgi?id=657990
This dramatically thins down and sanitizes the application code.
The ShellAppSystem changes in a number of ways:
* Preferences are special cased more explicitly; they aren't apps,
they're shortcuts for an app), and we don't have many of them, so
don't need e.g. the optimizations in ShellAppSystem for searching.
* get_app() changes to lookup_app() and returns null if an app isn't
found. The semantics where it tried to find the .desktop file
if we didn't know about it were just broken; I am pretty sure no
caller needs this, and if they do we'll fix them.
* ShellAppSystem maintains two indexes on apps (by desktop file id
and by GMenuTreeEntry), but is no longer in the business of
dealing with GMenuTree as far as hierarchy and categories go. That
is moved up into js/ui/appDisplay.js. Actually, it flattens both
apps and settings.
Also, ShellWindowTracker is now the sole reference-owner for
window-backed apps. We still do the weird "window:0x1234beef" id
for these apps, but a reference is not stored in ShellAppSystem.
The js/ui/appDisplay.js code is rewritten, and sucks a lot less.
Variable names are clearer:
_apps -> _appIcons
_filterApp -> _visibleApps
_filters -> _categoryBox
Similarly for function names. We no longer call (for every app) a
recursive lookup in GMenuTree to see if it's in a particular section
on every category switch; it's all cached.
NOTE - this intentionally reverts the incremental loading code from
commit 7813c5b93f. It's fast enough
here without that.
https://bugzilla.gnome.org/show_bug.cgi?id=648149
When porting to the new gnome-menus API in commit 8f3bdd4f1, the
initial loading of settings apps was left out, so settings panels
are neither found nor can be launched from the top panel menus.
settings.menu was removed in gnome-menus commit
b68bcd27f44ce2c494f6e3cd9695890b9c02af04; gnomecc.menu is the intended
replacement.
(On Red Hat Linux derived systems, settings.menu continues to exist)
https://bugzilla.gnome.org/show_bug.cgi?id=645063
In commit 9bd22dc0, I introduced an API to load an arbitrary
.desktop file, not necessarily from the menu path. It turns
out this function was broken because it created ShellApp instances
that were *different* from ones that were cached normally.
As far as I can tell, we didn't initially use it. Then later
Util.spawnDesktop was created which used this function.
Remove this broken function and all callers; if we're loading
.desktop files from *outside* the menu path, we can look at
readding.
This patch also kills off Util.spawnDesktop in favor of callers
talking to ShellAppSystem directly, now that the latter reports
errors.
https://bugzilla.gnome.org/show_bug.cgi?id=644402
Launch child processes more directly; we retrieve the PID, and
use it to keep track of the .desktop file we launched.
Now, when we get a window, since the X window has a PID, we
have a pretty strong association.
.desktop file <-> PID <-> window
And can thus map window back to .desktop file.
https://bugzilla.gnome.org/show_bug.cgi?id=637745
Add a "gicon" property so that a GIcon can be used instead of an
icon name, while still getting icon recoloring from the theme.
Also include a compatibility wrapper in libshell until GJS has
support for interface static methods.
https://bugzilla.gnome.org/show_bug.cgi?id=622451
StIconType will be used by a new StIcon class, so move it to the
header file of common enumerations. Including st-types.h which had
the St single-include check revealed that st-texture-cache.h didn't
have that check and several places were including that directly.
Fix that up.
https://bugzilla.gnome.org/show_bug.cgi?id=633865
* Use --warn-all, --warn-error
* Fix various broken gtk-doc
* Drop unused shell_get_event_related
* For header defines, we currently require them to end in _H to be skipped
* Drop the no-longer-necessary fix-meta-rectangle.py hack
* Move to the convention of using -private.h for headers that are,
well, private.
* Add shell-wm-private.h
The current search system uses the OR operator to concatenate search
terms. While results which are matched multiple times sort before
other matches, it is almost guaranteed that adding an additional term
to the search increments the number of results, which is rather
surprising.
https://bugzilla.gnome.org/show_bug.cgi?id=610955
Use GSettings for all Shell configuration. GConf is kept to read
configuration from external programs (Metacity, Nautilus and Magnifier),
but ShellGConf is removed because it's mostly useless for the few calls
we still have. Also get rid of unused GConf code in ShellAppSystem.
A basic GConf schema is still used to override Metacity defaults and
configure Magnifier in a system-wide fashion. GConf is also used as
GSettings backend via the GSETTINGS_BACKEND environment variable.
All of this will be removed when these programs have been ported
to GSettings and able to use dconf.
GLib 2.25.9 is required. Schemas are converted to the new XML format,
and compiled at build time in data/ so that the Shell can be run from
the source tree. This also requires setting the GSETTINGS_SCHEMA_DIR
environment variable both when running installed or from source tree,
in src/gnome-shell.in and src/gnome-shell-clock-preferences.in.
https://bugzilla.gnome.org/show_bug.cgi?id=617917
Separate out the main app view into different sections based on the categories
in the desktop file. The configuration is done via gmenu and the desktop menu
specification, we set XDG_MENU_PREFIX="gs-" on startup, so that gmenu reads
gs-applications.menu, which we install.
There is no support for "submenus" - only the menus directly under
Applications will be displayed as categories.
https://bugzilla.gnome.org/show_bug.cgi?id=614131
This patch combines several high level changes which are conceptually
independent but in practice rather intertwined.
* Add a "state" property to ShellApp which reflects whether it's
stopped, starting, or started. This will allow us to later clean
up all the callers that are using ".get_windows().length > 0" as
a proxy for this property
* Replace shell_app_launch with shell_app_activate and shell_app_open_new_window
A lot of code was calling .launch, but it's signficantly clearer
if we call this ".open_new_window()", and later if we gain the ability
to call into an application's menu, we can implement this correctly rather
than trying to update all .launch callers.
* Because ShellApp now has a "starting" state, rebase panel.js on top of
this so that when we get a startup-notification sequence for an app
and transition it to starting, it becomes the focus app, and panel.js
cleanly just tracks the focus app, rather than bouncing between SN
sequences. This removes display of non-app startup sequences, which
I consider an acceptable action in light of the committed changes
to startup-notification and GTK+.
https://bugzilla.gnome.org/show_bug.cgi?id=614755
The way we were loading data into a CoglTexture, then pulling it out
and manipulating it on the CPU, then loading it back into a texture
was a bit lame.
Clean things up a bit here by loading directly into the CPU, doing
the fading, then creating a texture.
Also cache the faded data in StTextureCache.
https://bugzilla.gnome.org/show_bug.cgi?id=612759
Brute force merge these two by essentially replacing St.TextureCache
with a (renamed) Shell.TextureCache.
One function was added for convenience, namely "st_texture_cache_load_file_simple".
St.TextureCache had a function to load a texture from a filename, and it
returned NULL on error but only half the callers actually checked this. This
function is better.
https://bugzilla.gnome.org/show_bug.cgi?id=607500
In case of duplicate infos structures with the same id, the
info structures we get from looking up the id in app_id_to_info
aren't necessarily the same as those we used to match, so we
can't rely on matching to implicitly initialize info->casefolded_name.
Since the name collation key isn't used in matching results,
just in sorting, init it on-demand in the sorting which is also more
efficient.
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 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
Clean up the vendor prefix handling a bit, and add "mozilla" so that
we pick up "mozilla-firefox.desktop" from Firefox's (recent?) change
to have a WM_CLASS of "Firefox".
This is a start at the "Active Appliction Item" component of the
shell design. Currently we just show the currently focused
application. When launching a new application, we show that as well.
The implementation here is not complete; basically when launching
we de-focus the active one, and the application well shows the
most recent startup sequence.
This kind of fails in the case of multiple sequences, and we
also don't correctly de-focus the current window in other
launch paths.
Pass the error variable to g_key_file_load_from_data_dirs in
Shell.AppSystem.get_default().load_from_desktop_file again, and
use a try/catch in places.js.
- Avoid error '"iconname" may be used uninitialized in this function'
by initializing said variable to NULL.
- Define shell_util_get_file_description as static (like the other
similar functions) to avoid another compiler error.
- Don't save errors from g_key_file_load_from_data_dirs into the
variable "error" (ie. pass NULL to it instead). Without this,
gnome-shell crashes if the key file can't be found (with message
"Error invoking Shell.load_from_desktop_file: Valid key file could
not be found in search dirs").
- Check the result of the load_from_desktop_file() call in places.js,
as it may be null.
Previously, ShellAppSystem only loaded (and cached) the set of
.desktop files from applications.menu and settings.menu, using
the gnome-menus library. The ShellAppInfo structure was
a "hidden typedef" for GMenuTreeEntry.
But we need to support loading an arbitrary .desktop file. Thus,
refactor the ShellAppInfo into a real struct, with a refcount,
and allow it to point to either a GMenuTreeEntry or a GKeyFile.
Also, in the case where we fail to lookup an icon for an
application, ensure we return a 0 opacity texture.
For both of these, because of optimizations a few patches ago, we
ended up relying on hash table ordering which caused instability
in the application well among other things. Define an ordering
for both.
The favorites is just the order of the GConf keys, and new items
get appended. In the future we should allow insertion at any
point which the grid could use.
For running applications order, define a new "initially_seen_sequence"
transient variable which is just an monotonically incrementing
integer assigned to an application for the first time we saw it
running in this session. When an application is closed, it's reset.
Before, we looked up application data in several ways; the ShellAppSystem
exported just application ids (though it parsed the .desktop files internally),
and we'd create a Gio.DesktopAppInfo object (reparsing the desktop file again),
wrapping that inside a JavaScript AppInfo class, and finally the AppDisplay
would again parse the .desktop file to get the categories.
Also, to look up applications by id previously, we traversed the entire
menu structure each time.
Some qualities such as the NoDisplay flag were not easily exposed in the old
system. And if we wanted to expose them we'd have to change several different
application information wrapper classes.
All in all, it was quite suboptimal.
The theme of this new code is basically "just use libgnome-menus". We do
not call into Gio for app lookups anymore. The new Shell.AppInfo class
is a disguised pointer for the GMenuTreeEntry item.
To fix the caching, we keep a simple hash table of desktop id -> ShellAppInfo.
Searching across NoDisplay desktop items can produce weird
results to the user (including duplicates, and items that
aren't really applications at all.) So, don't include them
normally.
But continue including NoDisplay items when we look up the
desktop file for a window, since we want to catch applications
like Evince and Nautilus which are otherwise NoDisplay.
http://bugzilla.gnome.org/show_bug.cgi?id=587548
Add a GConf key for favorites, and API for retrieving them.
Also add shell_app_system_lookup_basename, which we use from
the app monitor to look up WM_CLASS ids.
To avoid loading applications from two different systems, use
ShellAppSystem solely. This unifies the initial load and the
reload.
Extend ShellAppSystem to also load settings/preferences, and
ensure they appear in the search.
