When clicking the close button on a virt-viewer window with
a VNC session open, while the VNC session terminates, the
window does not go away.
The problem is that the virt_viewer_session_vnc_disconnected
method never gets invoked. The close button triggers a call
to virt_viewer_session_clear_displays which unrefs the
VirtViewerDisplayVnc instance. This in turn triggers a call
to gtk_container_destroy, which destroys all widgets it
contains, ie the VncDisplay * object.
With the VncDisplay object in its dispose phase, no signals
will ever be emitted, thus the 'vnc-disconnected' signal
never gets seen.
The design issue is that VirtViewerDisplayVnc is assuming
it owns the VncDisplay, whereas in fact the real owner is
the VirtViewerSessionVnc object.
The solution is to introduce a new virt_viewer_display_close
method which can be used to de-parent the widget before
VirtViewerDisplay is unref'd.
The VirtViewerSessionVnc object also needs to hold a full ref
on the VncDisplay object, not merely a floating reference
* virt-viewer-display-spice.c, virt-viewer-display.c,
virt-viewer-display.h: Add virt_viewer_display_close
* virt-viewer-display-vnc.c: Deparent VNC widget in
virt_viewer_display_close impl
* virt-viewer-session-vnc.c: Improve logging
* virt-viewer-session.c: Call virt_viewer_display_close
before unrefing display
* virt-viewer-window.c: Improve logging
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This has 2 advantages, and I can't figure any drawback:
- it fixes the issue of mnemonic hints being draw when pressing Alt
key (character underlined), even when they were disabled.
- it simplifies the code :)
This makefile is just fantastic, it forces you into good practices,
support various build targets (my windows builddir ignore the right
files etc..)
The more I use it, the more I like it.
The current code will attempt to dereference args if
--spice-controller, even if args is NULL.
Let's not accept any extra argument/uri on the command line if using
the controller. Beside, the conditionnal block looks better outside of
the if condition.
Lower warning message to debug level. There are various racy ways it
ends up calling show_display although the display is not yet
ready. This is not such a big problem, although it would be nice to
handle this case better
The current code only inform of focus state when the listener is ready.
spice-gtk controller code lacks signal when a client connects, but a
client will set the title when connected and send a notify signal.
Use this event to notify of application focus state.
Do not disconnect session when switching host (non-seamless migration
method).
Also, handle a bit better main channel events and do not disconnect on
unknown events, however raise unhandled event message to warning
level.
- auto-conf is an optionnal argument to --fullscreen:
it will set the guest display configuration to match the client
display configuration, by sending the client monitors size and
position to capable guests.
The current code notifies the controller when the remote-viewer
application starts, but not when the client is connected. We should do
the later instead
We require libvirt >= 0.9.7 to get virDomainOpenGraphics
We require spice-gtk >= 0.11 to get the fix for dealing with
authentication over an SSH tunnel
We requires spice-protocol >= 0.10.1 to get a constant
required by USB redirection
When running ./autogen.sh on a pristine git checkout, I got:
libtoolize: Consider adding `AC_CONFIG_MACRO_DIR([m4])' to configure.ac and
libtoolize: rerunning libtoolize, to keep the correct libtool macros in-tree.
You should add the contents of '/usr/share/aclocal/intltool.m4' to 'aclocal.m4'.
Trying to resize not visible windows leads to the following being printed
to the console:
Gdk-CRITICAL **: IA__gdk_window_get_origin: assertion `GDK_IS_WINDOW (window)'
This gets triggered by the gdk_screen_get_monitor_geometry() call in
virt_viewer_window_resize()
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Note this button only gets shown on USB redir capable virtual machines.
Changes in v2:
-Use gtk_widget_set_visible for simpler code
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
When quiting from the fullscreen menu call virt_viewer_app_quit instead of
gtk_main_quit so that the session gets properly disconnected before quiting.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Even though the previous patches in this series ensure that the session
gets properly finalized, we still need to wait for the disconnect signal,
as spice-glib uses co-routines which need some time to cleanly close the
connection / session.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Before this patch session-spice would emit the disconnected signal as soon
as the main channel is closed, but other channels may still be open at
that time and raising the disconnected signal usally leads to the app class
calling gtk_main_quit, at which point the other channels never get properly
finalized (as there co-routines still hold a reference to them).
This is esp. bad for usbredir channels as these re-attach the kernel driver
for redirected devices when finalized. So exiting without properly finalizing
them leads to the formerly redirected devices not being usuable until the
driver is manually reloaded or the device is unplugged and re-plugged
(the kernel does not automatically re-bind kernel drivers when userspace
closes a usbfs node).
This patch fixes this by delaying the emitting of the disconnect signal
until the last channel has been destroyed.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
With this patch combined with the previous patches in this series, the
VirtViewerSession (finally) gets properly finalized on exit.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
