In French, the singular form should be used for 0.

Christophe Fergeau authored Mar 12, 2014 and Zeeshan Ali committed Mar 12, 2014

sscanf man page says about %x:
'x      Matches an unsigned hexadecimal integer; the next pointer must be a
pointer to unsigned int'
get_decoded_udev_property() is using %02x, but this is not documented as
reducing the size of the expected pointer, it's only documented as causing
sscanf to read at most 2 digits of the parsed number.

Boxes was passing an uint8 to scanf instead of an uint, causing memory
corruption (plus a 'fortify: stack smashing detected' message if this is
enabled when building), and eventually a crash in get_decoded_udev_property().

There are two issues this patch fixes:

* g_file_native() thinks of paths to locally mounted remote locations, as
  non-native so Boxes errors out if given such paths. This patch fixes
  the issue by replacing g_file_native() usage with a mannual check.

* g_file_get_uri() gives us the remote URI even if you created the GFile
  argument using the local path of the remote file. This patch fixes the
  issue by using the original commandline argument if the argument is
  not a path rathan than URI.

Also we restrict the path/URI to filesystems that we know libvirt/qemu
can handle: local and mounted SMB shares.

https://bugzilla.gnome.org/show_bug.cgi?id=688798

https://bugzilla.gnome.org/show_bug.cgi?id=688798

In commit 3c1b44c0, we lost the ability to handle local paths in wizard.
This patch brings back that ability.

https://bugzilla.gnome.org/show_bug.cgi?id=688798

Boxes would just hang in wizard's preparation stage if you pass a path
to a file on a locally mounted remote share (e.g SMB). Although the real
solution is to be able to handle those, we should have this in place
just in case this codepath is walked for some other reason.

https://bugzilla.gnome.org/show_bug.cgi?id=688798

As strongly suggested by ebassi on IRC:

<ebassi> [19:34:50] zeenix: and I would *strongly* suggest you use a
proper ClutterTransition in this case, instead of relying on implicit
animations, since you're controlling the start and end states, and you
want to get notifications

Besides being the right thing to do here, this also works around the issue of
Boxes unable to change the actor visibility at the end of transition and
avoid the warning on console about transition being null against clutter
1.16 in certain cases.

https://bugzilla.gnome.org/show_bug.cgi?id=700306

Conflicts:
	src/util-app.vala

"left" was translated as "droit" rather than "gauche"

Bug reported by Arnaud Bonatti.

https://bugzilla.gnome.org/show_bug.cgi?id=723057

Yosef Or Boczko authored Jun 20, 2013 and Zeeshan Ali committed Jul 29, 2013

https://bugzilla.gnome.org/show_bug.cgi?id=702759

We were setting the label to ellipsize but without setting max-width on
labels, this was useless.

https://bugzilla.gnome.org/show_bug.cgi?id=690757

https://bugzilla.gnome.org/show_bug.cgi?id=690757

Before starting an opacity animation, ensure that there isn't one
already scheduled.

This patch fixes at least two cases of actors not showing-up:

1. When right-clicking on an item in collection view: overlay bar
doesn't become visible.
2. When launching Boxes with installer media argument: you end up in the
wizard but with main wizard area invisible.

https://bugzilla.gnome.org/show_bug.cgi?id=701635

Christophe Fergeau authored Jun 06, 2013 and Zeeshan Ali committed Jul 29, 2013

When an error occurs during SPICE connection, a channel-event event
is emitted, and its type is SPICE_CHANNEL_ERROR_XXX. Boxes currently
catches some channel-event events, but does not do anything in case
of errors. When such an error occurs (for example because of a wrong
remote host certificate), Boxes UI is stuck on a centered half zoomed
screen, and does not do anything.
By emitting the disconnected() signal when such an error is caught, we
ensure we will go back to the collection view.

https://bugzilla.gnome.org/show_bug.cgi?id=701753

Christophe Fergeau authored Jun 06, 2013 and Zeeshan Ali committed Jul 29, 2013

At the end of installations, regardless of if they are express
installs or not, setup_post_install_domain_config is called.
It unconditionally tries to remove 'disk_file' from the VM configuration.
However, this member is only set for express installs, it's NULL
for non-express installs, so we get glib criticals in the console
when trying to use it as if it was non-NULL.
This commit avoids this issue by checking if it's NULL before using
it, as is done in most of unattended-installer.vala

https://bugzilla.gnome.org/show_bug.cgi?id=701695

Christophe Fergeau authored Jun 05, 2013 and Zeeshan Ali committed Jul 29, 2013

The fix for bug #683489 changed the various VM sources to be loaded
sequentially. As some sources can be slower than others (remote oVirt
VS local libvirt connection), this means a slow source could block
processing of a faster source, causing the UI to be empty for longer
than it should.

This commit follows one of Zeeshan's suggestion to make that processing
asynchronous again.

Christophe Fergeau authored Jun 05, 2013 and Zeeshan Ali committed Jul 29, 2013

When authentication to an oVirt broker is cancelled,
yield OvirtBroker::add_source() would never return until the
authentication times out. Together with the changes
6b95a50c 'app: Avoid adding duplicate sources', this could cause
the UI to appear empty for the duration of the timeout if an oVirt
source is created first, and if the user cancelled the connection
(or even if the connection takes quite some time).

https://bugzilla.gnome.org/show_bug.cgi?id=683489