Re: [PATCH] Paravirt framebuffer backend tools [2/5]

[Anthony Liguori <aliguori@cs.utexas.edu>]

Jeremy Katz wrote:
> On Mon, 2006-09-04 at 10:01 +0100, Steven Smith wrote:
>   
>>> +CFLAGS += -g -Wall
>>>       
>> You shouldn't need to add -g here; Rules.mk handles it for you if
>> debug is set.
>>     
>
> *nod*  -Wall gets set in Config.mk as well -- will nuke.
>
>   
>>> --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
>>> +++ b/tools/xenfb/keymapping.c	Sat Sep 02 15:19:25 2006 -0400
>>> @@ -0,0 +1,141 @@
>>> +#include <stdint.h>
>>> +#include <gdk/gdkkeysyms.h>
>>> +#include <linux/input.h>
>>> +
>>> +uint32_t gdk_linux_mapping[0x10000] = {
>>> +	[GDK_a] = KEY_A,
>>>       
>> This is kind of ugly.  Is there any chance it could be autogenerated?
>> Also, where did 0x10000 come from?
>>
>> Also, depending on GTK just for the keymap table is a real pain.  Or
>> is it already required for libvncserver?
>>     
>
> libvncserver requires GTK.  And I don't know that there's really any
> good way to auto-generate it unfortunately.  I somehow expect that
> 0x10000 came from "it'll be big enough" but Anthony would have to
> confirm :-)
>   

That's the biggest that a GDK scan code can currently be.  That way, we 
can use a simple indexed table.

Regards,

Anthony Liguori

> The mappings are unfortunately a bit of a fact of life since we have to
> convert from what the X layer gets to what the kernel expects.  And the
> two couldn't be farther from the same.  And then it's even more fun when
> toolkits get involved.
>