[Spice-devel] [PATCH spice-gtk 0/5] Require GStreamer, fix build warnings

[Marc-André Lureau]

Hi

On Mon, Jan 7, 2019 at 2:50 PM Christophe Fergeau <cfergeau at redhat.com> wrote:
>
> On Mon, Jan 07, 2019 at 02:12:34PM +0400, Marc-André Lureau wrote:
> > Hi
> >
> > On Mon, Jan 7, 2019 at 2:03 PM Christophe Fergeau <cfergeau at redhat.com> wrote:
> > >
> > > On Sun, Jan 06, 2019 at 12:24:33AM +0400, Marc-André Lureau wrote:
> > > > >
> > > > > > I'll work on a patch to remove pulse.
> > > > > >
> > > > >
> > > > > It seems too soon.
> > > >
> > > > yes, we haven't done enough testing of the gstreamer backend on Linux.
> > > > The recording path at least fails very often for me due to a race:
> > > > https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/merge_requests/69
> > > >
> > > > After fixing the pulsesrc element, I added a check in spice-gtk for
> > > > pulseaudio plugin version. I don't see what else I could do.
> > > >
> > > > I propose to make GStreamer audio backend the default, and deprecate
> > > > the pulse backend.
> > > >
> > > > If the gst pulseaudio plugin is too old (<1.15), then fallback on
> > > > spice-gtk pulse backend for now.
> > >
> > > In other words, all systems using a stable release of gstreamer (latest
> > > is 1.14.4) will be falling back to the pulse backend for now, but we'll
> > > still pretend the pulse backend is deprecated and should not be used.
> > > And even if distros patch their 1.14 pulsesrc, we won't detect it and
> > > still fallback. In my opinion, it's unfortunately too early to be too
> > > that pushy on the switch, at least with respect to deprecation warnings.
> >
> > If we don't expose it as default, it will be hard to find problems
> > though. The pulsesrc bug is present since 2011-11-24..
>
> Yeah, I'm only questioning how verbose we should be when gstreamer is
> not used. Iirc the patch detecting if pulsesrc is too old outputs a
> g_warning if it is, there's a patch adding a warning if
> --with-pulseaudio is used, ... Imo at least the first one should only be
> a g_debug() for now.

g_warning or g_debug, not a big difference to me here.

But g_debug() is likely to be missed, since it is silent by default.

For the build-time warning, I think it is fair since it is mostly on
the packager side, and they should prepare themself and update
gstreamer.

The gst fix is queued for 1.14.5 now. I will update the patch checking
the version, and update the build warning to mention runtime
"requirement" of gstreamer 1.14.5


>
> Christophe
>
> >
> > Otoh, it's not such a bad bug either & the GStreamer backend works
> > fine on Windows for a long time.
> >
> > The GStreamer fix should be backported to -stable releases. I'll
> > discuss this with upstream GStreamer.
> >
> > --
> > Marc-André Lureau
> > _______________________________________________
> > Spice-devel mailing list
> > Spice-devel at lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/spice-devel



-- 
Marc-André Lureau