[Spice-devel] keypress-delay issue

[Marc-André Lureau]

----- Original Message -----
> 
> 
> On 06/23/2015 02:23 AM, Jared Kwek wrote:
> > Hi all,
> >
> > I worked directly with Victor to test out his patch and wanted to
> > report back to the mailing list on my results.  I set
> >  SPICE_KEYPRESS_DELAY=10 and saw an improvement in the response time
> > when I was typing.  I was using Fedora 21 in the client with Gnome 3
> > and tried typing in LibreOffice Writer and in the Gmail web mail
> > client on Firefox on my Fedora 22 VM as the SPICE server.  The lag is
> > somewhat noticeable with the keypress delay set to 100, such that when
> > I compare typing on a physical machine vs a SPICE session I can
> > definitely feel the difference.  I say somewhat noticeable because it
> > is not a huge difference; I can just feel a lag when typing.  With the
> > 10 ms delay it felt more like real-time.
> >
> > I'm wondering where to go from here.  Victor did submit a patch for
> > the environment variable, so that will be available as a workaround
> > for those who know about it.  I guess it is up to you the developers
> > if you want to make configuration or fix something more mainstream.
> > There were a couple complaints about it on the mailing list that I dug
> > up, but it doesn't sound like there's been much public outcry to fix
> > it. Nevertheless, since I am presenting it as an issue, I'd be happy
> > to do any further testing if anyone is interested in looking into it a
> > bit further, as I would recommend it as something that should be
> > addressed.
> >
> I believe I was one of those reports you dug up.  I had originally
> compiled in the changed delay and found the experience much more
> pleasant ; nothing is broken the way it is now, but the feel is
> significantly improved with the reduced latency.
> 
> I would like a configurable for this setting, and it is implemented, as
> I recall, as a widget property so at least at some level it is
> configurable.
> 
> The environment variable approach that's floating would work but is so
> hidden that the two of us would be the only ones on the planet using
> it!  But aside from that, I'm all for it!

It's interesting that I can barely notice the lag, and possibly most
people, judging by the amount of people complaining.

This was introduced to solve a real problem though, to avoid spurious
software key repeat on guest side. This is particularly annoying with
remote guest with a lag of several 100's of ms, but could also happen
locally if the VM get stuck due to scheduling.

I am not sure more tweaking should be advertised, as this is fairly
obscure and could reopen the issue I was trying to solve. Instead I
wish we would have a better/more clever solution to this issue.
Unfortunately, I don't have good one, adjusting the latency
based on local vs remote "measurements" would still let spurious
key repeats on slow scheduling, is this a better tradeoff?