[Spice-devel] [vdagent-linux 1/2] randr: remove monitors.xml on auto-configuration

[Marc-André Lureau]

On Wed, Oct 8, 2014 at 1:11 PM, Christophe Fergeau <cfergeau at redhat.com>
wrote:

> On Wed, Oct 08, 2014 at 12:28:53PM +0200, Marc-André Lureau wrote:
> > Hi
> >
> > On Thu, Oct 2, 2014 at 2:10 PM, Christophe Fergeau <cfergeau at redhat.com>
> > wrote:
> >
> > > Hey,
> > >
> > > On Wed, Aug 27, 2014 at 07:22:07PM +0200, Marc-André Lureau wrote:
> > > > From: Marc-Andre Lureau <marcandre.lureau at redhat.com>
> > > >
> > > > GNOME will restore monitors.xml configuration whenever the timestamp
> > > > "config > change". The "change" timestamp is the last user applied
> > > > configuration, whereas the "config" timestamp is updated when
> > > > the screen is updated or ouput/crtc modes are added/removed.
> > > >
> > > > These condition are triggered by vdagent during monitor config.
> Since we
> > > > can't control the timestamps (playing with delay will be inherently
> > > > event more racy), the only sane way I can think of is to disable gsd
> > > > behaviour. This can be achieved by deleting the
> ~/.config/monitors.xml,
> > > > which is the intended configuration to restore, so vdagent will
> override
> > > > whatever configuration was saved previously.
> > > >
> > > > Somehow, if vdagent would be better integrated with gnome2, it would
> use
> > > > the gnome-rr and/or org.gnome.SettingsDaemon.XRANDR dbus
> > > > API. Thanksfully, in gnome3, the monitor auto-configuration has been
> > > > merged in.
> > >
> > > Actually a bit curious how this relates to
> > > https://bugzilla.gnome.org/show_bug.cgi?id=706735 which seems to have
> > > added a hack to avoid similar situations ?
> > >
> > >
> > There is a similar timestamp check in gsd. However, when enabling
> monitors
> > with vdagent, the timestamps are very close and there is a race between
> > vdagent & gsd: you get random results. With the proposed patch, spice
> > client = vdagent wins over gsd when it wants to set some config.
>
> You make it sound like it's only an issue when running GNOME2. If you
>

I already mentionned recent gnome3 auto-config has been merged in (last
year)  and don't reach the issue (no related code).

tested with gnome-settings-daemon in RHEL6, the relevant (if I looked in
> the right place!) code seems different from upstream because of a
> downstream patch.
>

You are correct, showing that this part of gsd code doesn't have a clean
solution and needs various workarounds.


> I also assume it will not be an issue either when
> monitor configuration is not done through the agent but through
> qxl/client monitor config?
>

What do you mean? Monitor are only enabled through vdagent here.


>
> If all of that is correct, and given that we don't own monitors.xml, I'd
> prefer not to have this patch upstream.
>

This patch does not harm, we don't want gsd to race with vdagent: vdagent
should win over for auto-conf/resize to work properly.

-- 
Marc-André Lureau
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/spice-devel/attachments/20141008/d2ca976c/attachment.html>