[Spice-devel] [patch 0/2] vdagent KEYVAL extension

[Marc-André Lureau]

----- Original Message -----
> > Yes, although I think we should be able to send arbitrary utf8 input in the
> > protocol. If I read your proposal right, you are sending UCS-4 codes, no
> > utf8?
> > 
> > I think you should use gdk_keyval_to_unicode() then g_ucs4_to_utf8() for
> > key
> > input and efficiency (in your spice-gtk patch).
> 
> Unfortunately, the unicode people forgot to add values for function keys and
> other special
> keys! That is why I use the X11 keysmys - they can also represent all keys on
> the keyboard.
> AFAIK VNC protocol also use that.

Keys that don't have utf8 equivalent should be sent with the current input, I don't see a benefit changing that.

Having a utf8 input is mainly useful for utf8 input from client (input method, browser) and to bypass guest keymaps.

It shouldn't be used for all inputs.

> 
> > Do not send both to agent and hw
> > input layer, because the agent can't easily tell which of these two events
> > should
> > be discarded. The client however, should know when the agent fully supports
> > utf8 input. However, this will probably be more difficult to handle in the
> > agent,
> > when no input method is available (console etc). Either the utf8 input will
> > need
> > to be translated and sent to uinput, but that's not very elegant, or the
> > agent will
> > have to inform the client to switch to hw input.
> > I guess this is not a concern for spiceterm, because it is implementing an
> > agent
> > itself,
> 
> yes
> 
> > and cannot be turned off, and also because you handle all inputs. Perhaps
> > you could introduce spiceterm? The repo is lacking a simple README file
> > just to
> > make things clear what is really about :)
> 
> OK, will send more info tomorrow.
>  
> 
> _______________________________________________
> Spice-devel mailing list
> Spice-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/spice-devel
>