[Spice-devel] [PATCH spice-gtk] widget: fix request_auto_usbredir() critical

[Frediano Ziglio]

> 
> Hi,
> 
> On Wed, Jun 06, 2018 at 06:19:27PM +0200, Marc-André Lureau wrote:
> > On Wed, Jun 6, 2018 at 6:03 PM, Victor Toso <victortoso at redhat.com> wrote:
> > > On Wed, Jun 06, 2018 at 05:23:24PM +0200, marcandre.lureau at redhat.com
> > > wrote:
> > >> From: Marc-André Lureau <marcandre.lureau at redhat.com>
> > >>
> > >> On f28, when focusing out of the display, we get the following
> > >> critical:
> > >>
> > >> (spicy:15388): GSpice-CRITICAL **: 17:20:07.710:
> > >> spice_gtk_session_request_auto_usbredir: assertion
> > >> 's->auto_usbredir_reqs > 0' failed
> > >>
> > >> This is due to unbalanced gtk+ focus-in and focus-out events (one more
> > >> focus-out). This may be fixable on the gtk+ side, but it's also easy
> > >> to prevent on our side when the last focus state is unchanged.
> > >>
> > >> Signed-off-by: Marc-André Lureau <marcandre.lureau at redhat.com>
> > >
> > > Have you seen https://gitlab.gnome.org/GNOME/gtk/issues/792 ?
> 
> > Nope, thanks for the link.
> >
> > >
> > > IMHO it can be prevented in spice-gtk but some warning should be
> > > kept as this is unexpected.
> >
> > The function looks generic enough to prevent doing unnecessary
> > work if the state is not changed.
> >
> > I am not sure it's spice-gtk job to report gtk+ bugs :)
> 
> Sadly we rely on gtk and gtk bugs are easier to track if we have
> some info on it. The critical that you are fixing is bad, could
> be avoided by this patch but this patch workaround a bug in gtk+
> so we should know about it somehow with some debug/warning
> message, imho.
> 
> > How long before the gtk+ bug is fixed (if it's ever fixed)?
> 
> The fix was pushed [0] but need to double check if it was pushed
> to the stable branches, etc.
> 
> [0] https://gitlab.gnome.org/GNOME/gtk/commit/d3885e92a7db130d96865
> 
> > I'd just go with this simple patch, and let the gtk+ team
> > decide how and when to fix the gtk+ behaviour.
> 
> I'm not saying no for your patch, I prefer warning/debug message
> due gtk+ event bug instead of critical message later on.
> 
>         toso
> 

Is it only me or this patch is fixing an usb error changing
focus handling?
Won't be better to have also a similar check in 
spice_gtk_session_request_auto_usbredir ?

like:

    if (state == s->auto_usbredir_reqs) {
        return;
    }
    s->auto_usbredir_reqs = state;

so on first TRUE will change, on second/third/so on will ignore,
first disable will disable, other disable ignore?

> > >
> > >> ---
> > >>  src/spice-widget.c | 3 +++
> > >>  1 file changed, 3 insertions(+)
> > >>
> > >> diff --git a/src/spice-widget.c b/src/spice-widget.c
> > >> index 72f5334..2adcf30 100644
> > >> --- a/src/spice-widget.c
> > >> +++ b/src/spice-widget.c
> > >> @@ -207,6 +207,9 @@ static void update_keyboard_focus(SpiceDisplay
> > >> *display, gboolean state)
> > >>  {
> > >>      SpiceDisplayPrivate *d = display->priv;
> > >>
> > >> +    if (d->keyboard_have_focus == state)
> > >> +        return;
> > >> +
> > >>      d->keyboard_have_focus = state;
> > >>      spice_gtk_session_set_keyboard_has_focus(d->gtk_session, state);
> > >>
> > >>
> > >> base-commit: 7b67ed4eb6ab530bf58afac8ff4ed190cc951dfb