We were calling twice showPage() with the correct page, here and in
show() / zoomFromOverview given that _resetShowAppsbutton was called
from the signal 'showing' of overview. Given that the call to
_resetShowAppsbutton is only actually used when hiding the overview we
can actually put the checked state of the button to false when animating
from overview so it shows the workspace page, causing the same behavior
of _resetShowAppsbutton without all the shenanigans of resetting when
the hiding overview signal is triggered.
https://bugzilla.gnome.org/show_bug.cgi?id=732901
Currently the indicators are a BoxLayout inside a BinLayout in AllView.
BinLayout doesn't have any size constraint, so if the indicators request
a bigger size than AllView the entire overview is grown, causing the
overview to go crazy.
To avoid that, create an actor for the page indicators that request as
minimum size 0, and as a natural size, the sum of all indicators natural
sizes. Then we clip_to_allocation, so it doesn't grow more than the
parent.
https://bugzilla.gnome.org/show_bug.cgi?id=723496
If the application reports itself as single window (through
an explicit indication in the desktop file or some heuristics),
not show a "New window" item that doesn't actually open a new window.
https://bugzilla.gnome.org/show_bug.cgi?id=722554
Both Panel.ActivitiesButton and its parent class Panel.MenuButton would
attempt to connect their own _onEvent() function to Clutter::event,
which counterintuitively was connecting the child class' _onEvent()
function twice.
So, actually chain up on the signal handler, and don't connect twice
to the signal. Both methods were calling this.menu.close(), so only
do that on the parent class handler, since we're chaining up and doing
the right thing now.
https://bugzilla.gnome.org/show_bug.cgi?id=733840
We don't need to wait to until the stage window is mapped to take
the modal grab, because that code now runs in a startup-prepared
signal handler, which in turn runs some time after the mainloop
has started and well after the stage window is mapped.
https://bugzilla.gnome.org/show_bug.cgi?id=711682
The grab would previously just consume the button release, while propagating
motion events, possibly down to clients in wayland. This would produce
inconsistent streams there.
On pointer events, the inconsistency would just be having clients receiving
events with the button 1 set in the mask, with no implicit grab. When touch
events are handled, this would be more hindering as the client would receive
touch_motion events with no prior touch_down nor later touch_up, something
never supposed to happen.
https://bugzilla.gnome.org/show_bug.cgi?id=733633
The animation is the same for modal dialogs, but it is now
run for non modal dialogs too (matching the new behavior on
show).
In addition, we run a destroy animation for normal windows,
if they use CSD (there are technical limitations that prevent
running animations after destroy on server decorated windows)
https://bugzilla.gnome.org/show_bug.cgi?id=732857
Handle touch events, so that an interacted button locks to a single sequence,
but multiple sequences are free to interact with multiple key buttons.
https://bugzilla.gnome.org/show_bug.cgi?id=733633
The long press code has been refactored so it can be used on both pointer and
touch events, and the click gesture has been made to account for button=0.
https://bugzilla.gnome.org/show_bug.cgi?id=733633
No sequence checks are done, these UI elements promptly trigger a grab that
will cancel ongoing touches and redirect later ones somewhere else, so that
works as a barrier to multi-toggling.
https://bugzilla.gnome.org/show_bug.cgi?id=733633
This adds a table with mappings for GNOME apps that have recently
renamed their desktop files, and uses that to update the desktop names
saved in user settings with the new values.
https://bugzilla.gnome.org/show_bug.cgi?id=729429
Currently to know how many results we could show for GridResults
we use the width of the bin containing those results. Since it's
expanding it shouldn't be a problem. But it becomes a problem when
no results are displayed, thus the container becomes hidden and
it losts its allocation.
In the next introduction of terms in search we call again
maxDisplayedResults but it doesn't have allocation yet, and therefore no
results are displayed (currently a bug on IconGrid makes the min size =
one icon, so actually we show one and only one icon in this case).
To solve that use the parent container which contains the search results
of all providers or the text label with not displayed results, so it
always have the real available width to calculate maxDisplayedResults.
Thanks Alban Browaeys for the debugging footwork.
https://bugzilla.gnome.org/show_bug.cgi?id=732416
Unlike for the main app view, where we only move the key focus once the
users starts navigating, the key focus is moved immediately when opening
a folder popup. This is unexpected, so make app folders consistent with
the main view.
As arrow keys will not work while the container itself has key focus, we
handle those explicitly by translating them to TAB_FORWARD and
TAB_BACKWARD respectively.
https://bugzilla.gnome.org/show_bug.cgi?id=731477
Support was added to Mutter to allow it to trigger a restart
to allow for restarts when switching in or out of stereo mode.
Hook up to the new signals on MetaDisplay to show the restart
message and reexec. Meta.is_restart() is used to suppress
the startup animation.
This also allows us to do 'Alt-F2 r' restarts more cleanly
without a visual flash and animation.
https://bugzilla.gnome.org/show_bug.cgi?id=733026
Instead of always logging frame timestamps for every frame - which
was using >26 bytes of memory per frame, or 5MB per hour of continuous
redrawing - make frame timestamps something that defaults off and is
turned turned on using a new ShellGlobal::frame-timestamps property by
the perf scripts.
https://bugzilla.gnome.org/show_bug.cgi?id=732350
Add an option for windows created with Scripting.createTestWindow()
to continually redraw themselves; this is for testing performance
of application updates.
https://bugzilla.gnome.org/show_bug.cgi?id=732350
We currently replay events that should start a search to the search
entry, which is fairly dodgy. Synthesize a new event with the correct
source actor instead, which is a bit less evil.
If an active grab on pointer events happens during multi-touch operations,
all non-pointer-emulating touches will be muted. This may leave the
Clutter.ClickAction incomplete if triggered by one of those sequences,
just to have a gesture take over and trigger a compositor grab, which would
leave the capture-event handler stuck eating events.
So listen for grab-op-begin from the display, and ensure the action is
released if such grab begins.
and the capture event handler stuck.
When returning to the desktop from overview we always show the
workspaceDisplay, given that is which have the windows clones to allow
animations.
The problem becomes when previous that we were at some other
page, like Search or AppDisplay. The problem is that when showing the
workspaceDisplay the windows are repositioned. That's wanted except
when returning from overview, since that causes unwanted animations
of the windows.
To avoid that just not reposition the windows if leaving the overview.
We don't normally hit the code in scripting.js to print metrics
because shell-perf-tool bypasses it, but there was a left-over
in the code that no longer works. Also add in the units to the
output.
https://bugzilla.gnome.org/show_bug.cgi?id=732349
If reusing an existing separate builddir, the generation of dependencies
for the JS resource file would fail because it couldn't find misc/config.js,
resulting in the resource file not being regenerated.
https://bugzilla.gnome.org/show_bug.cgi?id=732348
Listen to changes in connectivity, and ask our helper to authenticate
when needed.
We don't have a URL to connect to yet (we will have when
the new NM API lands), so we use the default of trying
www.gnome.org (which is also more reliable because we can
recognize when the login is done)
https://bugzilla.gnome.org/show_bug.cgi?id=704416
Add a small DBus-activated GtkApplication that embeds a WebKitWebView
and implements some minimal logic to see if the login succeeds.
It will try to connect to a custom NM-provided url (the portal login
page), if one exists, or to www.gnome.org in the normal case of
a portal doing redirect.
https://bugzilla.gnome.org/show_bug.cgi?id=704416
Icons inside the menu are updated only for device state change,
but for the main device they also depend on connectivity (which
is a global property).
Add a public method to force an update of the icon, and call it
when connectivity changes.
https://bugzilla.gnome.org/show_bug.cgi?id=726401
They are different properties, they deserve different syncs.
Especially because a full allocation cycle sets both anyway, so
we should save some cycles this way.
https://bugzilla.gnome.org/show_bug.cgi?id=729823
We already have the width and height information cached in JS,
let's avoid going through gjs-gobject-clutter to retrieve them
again. As a plus, with normal properties the optimizer should
be able to generate better code.
https://bugzilla.gnome.org/show_bug.cgi?id=729823
We must remove the old image from the cache, not the new one.
This was causing a leak of old (and expensive) background
images, and was causing errors at the end of animations, trying
to destroy the animated background.
We translate 'On' to available accuracy level but if available accuracy
level later changes, we don't update available accuracy level accordingly
and hence limit the accuracy of apps.
E.g if available accuracy level is 'city' and geolocation is enabled,
the max accuracy level would be 'city' and apps can't get higher than
that. Now if user plugs in GPS, the available accuracy level will change
to 'exact' but without this patch max accuracy level will remain to be
'city' and apps will not be able to use the GPS.
https://bugzilla.gnome.org/show_bug.cgi?id=731882
The code currently tries to use Meta.KeyBindingFlags.REVERSED. Since
this constant is |'ed with Meta.KeyBindingFlags.REVERSES, gjs silently
ignores the unknown flag.
https://bugzilla.gnome.org/show_bug.cgi?id=731619
Commit 5d00c1a5ee moved app folder popups to GrabHelper - for some
reason, the line that ensures the current behavior of only considering
events inside the app picker to dismiss popups got lost ...
As clicks outside the app picker should still be handled normally
while clicks inside should dismiss the popup, we cannot make full
use of GrabHelper. However using it at least for focus handling
fixes some minor details we are getting wrong, for instance not
restoring the previous focus after dismissing a folder popup.
Commit 6c2f3d1d17 moved pref overrides into JS to implement
session mode specific overrides in a clean and generic way.
However that approach comes with a cost - doing the overrides only
after having handled over control to JS means that the core will
be initialized with the non-overridden settings before changing
to the correct values. In the best case this is unnecessary work,
but it can in fact have a worse effect: when initializing workspaces,
we will restore the previous number of workspaces when using
dynamic-workspaces and reset to the configured number otherwise.
As the non-overridden default for dynamic-workspaces is FALSE, we
can easily end up moving the user's windows to the "wrong" workspace.
Now GSettings is expected to grow support for session specific defaults,
which will render our entire override system obsolete (yay!). Given
that, it seems acceptable to use a less generic (and uglier) approach
in the meanwhile, in order to fix aforementioned problems. So move
overrides back before core initialization and just hardcode the
session-mode => override-schema relation.
https://bugzilla.gnome.org/show_bug.cgi?id=695487
The 0x0 dummyCursor works well when the menu pops up directly underneath
the pointer (e.g. when triggered by right-clicking the titlebar) or by
keyboard, but not when triggered by the menu button - the menu does not
point to the center of the button's bottom edge, and unless the user
keeps holding the mouse button while moving into the menu, the menu will
be dismissed immediately on button release.
Address these issues by using the button geometry to overlay the window
button with an appropriately sized actor that acts as a proper sourceActor,
to make the window menu behavior consistent with other shell menus.
https://bugzilla.gnome.org/show_bug.cgi?id=731058
Having the full geometry of the menu's source button (if any) will
allow us to address several misbehaviors of window menus, so use
that instead of show_menu().
https://bugzilla.gnome.org/show_bug.cgi?id=731058
When the pointer leaves the notification area, we queue a timeout to
hide the notification after a little while. If the user is hovering over
a notification and clicks the X button to close the notification, we will
destroy the notification, which causes a "pointer left" event on the
notification area. This queues a timeout which erroneously fires after
the next notification in the queue shows up.
The code and state machine are too complex to properly make sure this
timeout doesn't fire when there is no notification up next, so instead
just clear it when showing a notification to make sure that any
previously queued timeout doesn't apply to us.
https://bugzilla.gnome.org/show_bug.cgi?id=731118
We don't make use of any functionality StTable provides over
ClutterTableLayout, so port all users to the Clutter layout
in order to remove our own copy of the code.
https://bugzilla.gnome.org/show_bug.cgi?id=703833
We don't make use of any functionality StTable provides over
ClutterTableLayout, so port all users to the Clutter layout
in order to remove our own copy of the code.
https://bugzilla.gnome.org/show_bug.cgi?id=703833
It is a bit odd to request AM/PM format when the locale selected
by LC_TIME lacks the concept. We ignore the format setting in that
case elsewhere and assume 24-hour format, let's do the same for
chat timestamps for consistency.
https://bugzilla.gnome.org/show_bug.cgi?id=728271
When workspaces are disabled by the session mode, all workspace
related actions don't make sense. Worse, the "Move to Workspace ..."
actions allow sending windows into a nirvana with no means to
get them back.
https://bugzilla.gnome.org/show_bug.cgi?id=730653
Starting keynav into the active page is handled from a key-press
handler on the stage, however we should not "start" keynav when
we are already navigating elsewhere - the latter can happen when
keynav fails (for instance because the focus is trapped inside an
open app folder or at the end of the dash), and the event bubbles
up to the stage. So make sure to only handle the event to actually
start keynav, to not interfere with the normal navigation handling.
Thanks to Carlos Soriano <carlos.soriano89@gmail.com> for the
debugging footwork.
https://bugzilla.gnome.org/show_bug.cgi?id=726760
The fallback app-menu in GTK+'s client side decorations obviously
uses the GTK+ theme rather than the shell one; update the style
of our own fallback app-menu to resemble that style.
https://bugzilla.gnome.org/show_bug.cgi?id=730752
We now allow "appmenu" in the button layout to make synchronizing it
with GTK+'s client-side decorations easier, but as some people tweak
their settings to get in-window app menus even when using the shell,
actually pop up the app menu when the button is activated.
https://bugzilla.gnome.org/show_bug.cgi?id=730752
The optional label support introduced in commit af063dc2f2 broke
the centering of separators, as the label still adds additional
spacing even when empty.
Properly hide the label actor in that case to fix the alignment.
https://bugzilla.gnome.org/show_bug.cgi?id=730753
The extension-prefs tool is used by gnome-tweak-tool and the
extensions web site to display preferences. However as those
already implement their own extension lists, the main window
is not useful in that context (to not say it is rather silly).
Just skip the main window and only show the specified extension's
preference dialog in those cases.
https://bugzilla.gnome.org/show_bug.cgi?id=730829
Bring the extension-prefs tool in line with the mockup by adding
switches to enable/disable extensions, similar to the extension
page in gnome-tweak-tool.
https://bugzilla.gnome.org/show_bug.cgi?id=730829
The extension-prefs UI has never been great, but as the GNOME 3
design patterns are evolving, it is starting to look seriously
outdated. Modernize the UI a bit to have it fit in a bit better.
https://bugzilla.gnome.org/show_bug.cgi?id=730829
When a monitor is removed, the OsdWindow for that monitor may process
the monitors-changed signal before OsdWindowManager does (which will
remove the OSD). If that happens, we will currently try to access
an invalid monitor; check for this to avoid a couple of warning.
When switching between alternatives in AltSwitcher, the currently
visible child is replaced with the alternative. If the original
child has the key focus when it is removed from the stage, the
focus is lost. Detect this case and manually move the focus to
the new child.
https://bugzilla.gnome.org/show_bug.cgi?id=727259
When dragging a window preview from a non-primary monitor onto a
workspace thumbnail, the window is moved to the primery monitor
first before changing its workspace. However when dragging the window
between thumbnails to create a new workspace, it is kept on its current
monitor instead. This is not only inconsistent, but outright confusing
with the default 'workspaces-only-on-primary' setting, as the newly
created workspace is immediately removed again.
https://bugzilla.gnome.org/show_bug.cgi?id=683819
We only need GSystem when running under systemd. As libgsystem itself
has a hard dependency on systemd, only import it when actually needed
to keep working on systems where systemd is not available.
https://bugzilla.gnome.org/show_bug.cgi?id=728449
The workspace switcher should be expanded when workspaces are in use.
Our current implementation assumes that workspaces are used when there
are windows on at least two workspaces. However workspaces are already
used when moving from a non-empty workspace to an empty one (presumably
with the intention to launch something on that workspace), so tweak the
heuristic accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=662457
Having the ability to go to the current date if the user is already
on the current date can be confusing. So don't make the button reactive
until the selected date changes.
https://bugzilla.gnome.org/show_bug.cgi?id=726724
Currently, our logic for page panning isn't great. If the user starts a
pan upwards and hesitates over a new page, we'll go to the *next* page
on release, since the difference is greater, but the velocity wound down
to 0.
Instead of trying to treat it like page down or scrolls, simply do the
math to find the page where the user scrolled to.
This is unfortunately broken for fast swipes, since the user doesn't get
far enough into the new page to make a difference. I'm getting the
impression we'll need a gesture recognizer for this, though, however
crude. Simple hacks I tried, like a velocity multiplier, didn't work
properly.
https://bugzilla.gnome.org/show_bug.cgi?id=729064
We often call goToPage like:
this.goToPage(this._currentPage - 1);
(or + 1). During panning, these are based on the velocity values
of the gesture action. If we're already on the first or last page
and the panning ends, it's possible to get goToPage that's either
-1 or greater than the last page.
During normal usage, this isn't a problem, since the Y values will
be correct, always. But when panning, the Y values stick to the
finger, and thus if we return early, we won't snap to the exact page,
making it seem like things get "stuck".
Fix this the simple way by clamping to the correctly-ranged values
of pageNumber
https://bugzilla.gnome.org/show_bug.cgi?id=729064
The order in which month and year are displayed is controlled by
a "special" translated string in GTK+. We pick up the translation
from there, so make sure that it doesn't get translated again
in gnome-shell.
https://bugzilla.gnome.org/show_bug.cgi?id=715042
When workspaces have been removed, we need to remove the corresponding
thumbnails as well; the number of thumbnails that need removing is
the difference between the old number of workspaces and the new one.
Currently we assume that the old number of workspaces corresponds to
the number of existing thumbnails, but that may actually be wrong:
A thumbnail will still be animated out after its workspace has been
removed. As a result, we end up removing too many thumbnails when a
workspace is removed while a thumbnail of a previously removed workspace
is still animating out. Fix this by basing the old number of workspaces
only on thumbnails that have not been removed previously.
https://bugzilla.gnome.org/show_bug.cgi?id=728820
Taking an area screenshot doesn't work currently when in "grab mode",
for instance when the message tray or top bar menus are open. Fix
this by using GrabHelper for selecting the area, so grabs are properly
stacked for us.
https://bugzilla.gnome.org/show_bug.cgi?id=709126
The code path is completely unmaintained and untested (and probably
unused as well, considering that nobody has complained about accessing
the session object's Active property which does not exist in the
ConsoleKit case).
Most of our ConsoleKit code is already a dummy anyway, just do the
same for the remaining functionality.
https://bugzilla.gnome.org/show_bug.cgi?id=686626
Commit 7101cc3170 caused a small
regression insomuch that it checks for a valid "level" but simply using
"if (level)" which will be false if level is undefined and if
level == 0.
Check for not undefined instead.
https://bugzilla.gnome.org/show_bug.cgi?id=727384
Sometimes it is more appropriate to set layout properties that are
hooked up to CSS properties in code. However this is currently not
possible, as we end up setting properties to 0 when not found in CSS;
be a bit more careful when hooking up CSS properties to support this.
https://bugzilla.gnome.org/show_bug.cgi?id=728897
When we unregistered providers, like when we refreshed the list of
active remote providers, we would forget to destroy the old provider
display after the fact. This left an empty "skeleton" provider display
still in the search results that would never be filled in. Make sure
to destroy it properly.
https://bugzilla.gnome.org/show_bug.cgi?id=728597
We had one osdWindow that we displayed either on the primary monitor or on
whatever monitor index got passed over dbus.
Change that to show the osd on all monitors when no explicit monitor is
requested. A monitor should be requested in cases like display brightness where it makes sense to only show the osd on the affected monitor.
https://bugzilla.gnome.org/show_bug.cgi?id=722684
NetWM struts are defined in terms of screen edges (rather than monitor
edges), which works poorly with vertical monitor layouts (as it renders
entire monitors unusable). Don't extend struts in those cases.
https://bugzilla.gnome.org/show_bug.cgi?id=663690
Calling Mainloop.source_remove() on the timeout ID will not reset it,
and we're already removing the timeout by returning GLib.SOURCE_REMOVE
in the callback.
If geoclue reports that we can't aquire location, we hide the menu. This
will typically happen when user is offline (and doesn't have a 3G
modem). This is likely not what we want since this like a temporary
situation and user would want the ability to toggle geolocation still
even if its currently not possible any applications to query the
location.
https://bugzilla.gnome.org/show_bug.cgi?id=727398
We need to use a GdkPixbufLoader instead of the straightforward
gdk_pixbuf_new_from_file(), since we want to load the image already
scaled if possible - e.g. if it's an SVG file.
https://bugzilla.gnome.org/show_bug.cgi?id=726907
* 'Turn On' -> 'Enable'
* 'Turn Off' -> 'Disable'
* 'Off' -> 'Disabled'
* 'On' -> 'In Use' or 'Enabled' depending on whether or not service is
in use.
https://bugzilla.gnome.org/show_bug.cgi?id=726498
Telepathy 1.0 will not be compatible, and will probably require
source changes. telepathy-glib 0.12 and telepathy-logger 0.2 are
the 0.x ABIs (they were the first stable-branches to have g-i).
Bug: https://bugzilla.gnome.org/show_bug.cgi?id=721704
Reviewed-by: Giovanni Campagna
When a user logs in to a wayland session, we keep the login screen
running on the X server with the login screen running in a deactivated mode.
This commit makes sure it get reactivated when the user comes back to
the VT (from user switching, logout or just ctrl-alt-f1).
https://bugzilla.gnome.org/show_bug.cgi?id=726989
In addition to BluetoothAirplaneMode, we need to check also
BluetoothHasAirplaneMode, which is indicative of bluetooth rfkill
devices (and by extension bluetooth adapters).
This prevents showing the menu if there is no adapter present.
https://bugzilla.gnome.org/show_bug.cgi?id=725057
Most system menu entries are disabled on the lock/login screen;
there is no good reason why users should be allowed to turn bluetooth
on/off (but not e.g. Wifi), so disable the entry as well.
https://bugzilla.gnome.org/show_bug.cgi?id=726319
When reacting to background settings changes, we may end up queuing
more than one load. The redundant backgrounds are expected to be
destroyed when the previous background has faded out; however since
commit 933f38390b, the tweened actor is the same for all
consecutive load operations and we end up with a single onComplete
handler, ergo a single destroyed actor.
As new backgrounds are always added to the bottom, we are not only
piling up additional background actors, but break changing backgrounds
more than once, as the correct background ends up being covered by
previously added redundant actors.
Fix this by destroying redundant actors right after loading rather
than waiting for the fade animation to complete.
https://bugzilla.gnome.org/show_bug.cgi?id=726120
Folders use a 2x2 grid of the first 4 app icons as icon - if a folder
contains less apps, we currently skip the corresponding grid positions.
As a result, the overall size request may be smaller than the one for
other icons, making the folder icon look out of place.
Fix this by always using a full grid as folder icon, using dummy actors
as necessary.
https://bugzilla.gnome.org/show_bug.cgi?id=726322
At some point ScreenShield had code to do this, I don't know when
it was lost, but it makes sense and avoids having to move the mouse
just to see the shield.
https://bugzilla.gnome.org/show_bug.cgi?id=726378
With the switch to a table layout in commit f959cafb36, setting
alignments to place the individual icons at the outer edge of the grid
stopped working. Remove that code and add some explicit spacing instead.
https://bugzilla.gnome.org/show_bug.cgi?id=726323
If the notification is updated while SHOWING, we'll overwrite the
tween updating it to the new 'y' position, but forget to update the
state to SHOWN at the end of our transition. Make sure to always set
the state to SHOWN at the end.
https://bugzilla.gnome.org/show_bug.cgi?id=704844
This code may have worked when written in 2009, but later gjs commit
b5e467d89aea43a8e32a1138d232c8a32e6b0785 removed the priority
parameter from idle_add.
Now, when running a constantly-updating client (es2gears) on an
embedded platform, _updateRegions() does not get called and I see
unresponsive window decorations.
Update the code to use meta_later_add() like other parts of the
shell, which is actually slightly better in this case anyway.
Solves the unresponsiveness problem.
https://bugzilla.gnome.org/show_bug.cgi?id=585500
The smartcard service is put in the foreground in two cases:
1) If password service is disabled by admin configuratoin
2) if a smartcard is inserted
In either case we don't want to ask the user to pick a user from the
userlist. We currently only avoid asking in case 2.
This commit fixes case 1.
https://bugzilla.gnome.org/show_bug.cgi?id=726263
We only want to react to the card the user logged in with, at
the unlock screen. We check "at the unlock screen" by checking
the "reauthenticating" state variable. That variable is the
wrong one, though. It gets set too late, and in some cases, gets
set at the login screen, too. We should be checking this._reauthOnly
instead.
This commit fixes that.
https://bugzilla.gnome.org/show_bug.cgi?id=726262
If a user inserts the smartcard they logged in with into the system,
it's supposed to lift the shield and prompt for pin. That doesn't
happen because the parameter list of the smartcard-inserted signal
handler is wrong.
This commit fixes that.
https://bugzilla.gnome.org/show_bug.cgi?id=726262
Doing so is inconsistent with the behavior in the summary, and
is quite annoying when dealing with chats (because there is no way
to unfocus a chat notification with the keyboard only)
https://bugzilla.gnome.org/show_bug.cgi?id=724178
We can't let live (ie, never destroyed) actors undergo GC, because
they will emit :destroy signals during finalization and assert/crash
libmozjs. Properly destroy all actors before letting the GC
free them.
https://bugzilla.gnome.org/show_bug.cgi?id=724798
If we have the following sequence:
cache.getImageContent({ filename: "foo", cancellable: cancellable1 });
cache.getImageContent({ filename: "foo", cancellable: cancellable2 });
cancellable1.cancel();
Then the second load will complete with "null" as its content, even though
it was never cancelled, and we'll see a blank image. Meanwhile, since the
second load simply appends to the list of callers for the second load,
cancellable2 does absolutely nothing: cancelling it won't stop the load,
and it will still receive onFinished handling.
To prevent this from happening, give the actual load operation its own
Gio.Cancellable, which is "ref-counted" -- only cancel it when all the other
possible callers cancel.
Based on work from Jasper St. Pierre <jstpierre@macheye.net>
https://bugzilla.gnome.org/show_bug.cgi?id=722149
At the moment when a file is loaded, we iterate through the list of
pending file loads and ignore any unrelated to the file, then iterate
all the callers of the related file loads and finish them.
In fact, there can only ever be one pending file load related to the
file, and we already know it, so we can avoid the ugly nested loops.
https://bugzilla.gnome.org/show_bug.cgi?id=722149
This commit moves the code around a bit such that the
caller gets allocated up front and then a file load is either
found or created to attach the caller to.
Functionally, the code is the same, it's just now factored in a way
that will make it easier to fix a bug with cancellation later.
https://bugzilla.gnome.org/show_bug.cgi?id=722149
