Re: [RFC] keymap support for PVFB

[Markus Armbruster <armbru@redhat.com>]

Gerd Hoffmann <kraxel@suse.de> writes:

>   Hi,
>
>>> I've tried to tackle the same issue by hacking the vnc client side to
>>> send us keysyms no matter what the local keyboard mapping is.  So I can
>>> have any keyboard map loaded on the host, qemu-dm/vncfb sees us keysyms
>>> nevertheless and passes the correct scancodes to the guest OS.
>> 
>> You mean scan codes, don't you?  Key symbols are the XK_a and so
>> forth.
>
> No, keysyms.  This is what the vnc protocol uses, so there is no way
> around that, unfortunaly.  It takes the X11 keycodes and translates
> these to us keymap keysyms using a buildin table, then sends them.  So
> for the server side (from vnc protocol view, i.e. qemu-dm or vnc-fb) it
> looks like a vnc client with us keyboard.
>
>> Passing scan codes in addition to key symbols makes sense.
>
> Does the vnc protocol allow that?  I don't think so :-(

Broken as designed...

Your hack has the advantage that you don't configure the client keymap
into the server, and therefore don't have to kick the server whenever
the client keymap changes.  But when your client receives a funky key
that ordinary US keyboards don't have, you don't have the key symbol
for it.

Hmm, can we make one up then?  I mean it's a hack to begin with, why
not put another hack on top of it?  As long as every key gets its own
key symbol, and the made-up ones don't screw up servers that don't
know of this hack...

>> Would you mind sharing your code when it's ready?
>
> http://www.suse.de/~kraxel/xen/, the vnc client in the "xenwatch" package.

Thanks.