This is caused by automake-1.15 changing the location of the build tree
during distcheck, and the fact that intltool doesn't ignore the path,
yet.

Original patch from Zbigniew Jędrzejewski-Szmek for systemd.

https://bugs.launchpad.net/intltool/+bug/1117944

Use the '-' prefix when including git.mk. Otherwise Automake will mark
git.mk as a dependency of the Makefile and pull it into the tarball.

It's been a year and we still don't know what exactly happened. Who
knows what is out there in the wild? So, instead of aborting let's try
to handle this more elegantly.

Keep the warning, remove the assertion and show an empty label.

https://bugzilla.gnome.org/show_bug.cgi?id=734766

https://bugzilla.gnome.org/show_bug.cgi?id=757015

https://bugzilla.gnome.org/show_bug.cgi?id=757015

https://bugzilla.gnome.org/show_bug.cgi?id=757199

A ViewModel instance can have "hidden" references held by the
GtkTreeRowReferences of each item that is present in the model. Since
items are owned by the ItemManager, they can outlive the Embed.

https://bugzilla.gnome.org/show_bug.cgi?id=757199

The whole point of updating the OffsetController's count is to ensure
that we can load more items and show the "No Results" page. If the
ViewModel is about to be destroyed, then the ViewContainer is also
gone, and the count is useless without a view. So we might as well
spare ourselves a needless query.

https://bugzilla.gnome.org/show_bug.cgi?id=757199

The button's icon was supposed to change colour when activated.
However, this broke recently, presumably after some changes to
Adwaita, and the icon won't retain the colour. Instead, it would
flicker for a bit and go back to the inactive colour.

https://bugzilla.gnome.org/show_bug.cgi?id=751106

We were losing the search results when turning off the searchbar before
moving to the preview. We avoid this by freezing the TrackerControllers
so that they don't react to any changes in the search parameters.

The TrackerControllers are thawed *after* we are back and the search
state has been restored to avoid needless queries.

https://bugzilla.gnome.org/show_bug.cgi?id=741593

... so that it is easier to freeze the TrackerControllers.

We don't want to call photos_tracker_controller_set_query_status when
reacting to PhotosOffsetController::offset-changed because the
ViewModels get cleared when the query status is TRUE. Let's handle this
by introducing a new flag.

https://bugzilla.gnome.org/show_bug.cgi?id=741593

This will let us freeze and thaw all the TrackerControllers without
having to manually maintain the list.

https://bugzilla.gnome.org/show_bug.cgi?id=741593

I came across a few intermittent failures to create the GomMiner
proxies while running goa-daemon under valgrind. Let's atleast handle
the failures and print the error message.

Now that we g_application_get_default instead of
photos_application_new, we don't drop the reference to the Application
instance. Hence this check is not needed.

Fall out from cf0c6199

Earlier, we had code in LoadMoreButton that would stop us from trying
to load more items that we actually had. We lost that when
LoadMoreButton was killed.

Since we no longer have all the code that looked at the vertical
GtkAdjustment and item count to decide whether or not to show the
button, we can directly bake this condition into
photos_offset_controller_increase_offset.

Fall out from 1d226d3d

https://bugzilla.gnome.org/show_bug.cgi?id=688566

https://bugzilla.gnome.org/show_bug.cgi?id=756221

Ensure that the fullscreen logic runs after priv->mode has been updated
so that there are no state changes between can-fullscreen-changed and
window-mode-changed.

This changes the current requirement that set_fullscreen has to be used
before the mode is updated, which I found to be confusing, and we don't
need a separate priv->can_fullscreen variable any more.

https://bugzilla.gnome.org/show_bug.cgi?id=756131

The return types are obviously gboolean, not PhotoWindowMode, even
though the C compiler doesn't notice the difference.

https://bugzilla.gnome.org/show_bug.cgi?id=756131

We forgot to do this, but the gnome-documents code got it right. So,
let's cherry-pick the change.

https://bugzilla.gnome.org/show_bug.cgi?id=756131

This keeps the code in sync with gnome-documents and will make it
easier to understand future changes.

https://bugzilla.gnome.org/show_bug.cgi?id=756131

The window manager can fullscreen us at any time. Make that situation
consistent with being fullscreened by our own action.

Original patch from Cosimo Cecchi for gnome-documents.

https://bugzilla.gnome.org/show_bug.cgi?id=725508

... by adding a matching call to gegl_exit in the shutdown method.

Original patch from Alessandro Bono for gnome-documents.

Fixes GdTaggedEntry theming and drops a deprecated property.

https://bugzilla.gnome.org/show_bug.cgi?id=746104