Rikki,

    May be... Wine uses the idea to map keys into vkeys and then mapping back to keys... If something goes wrong on this proccess, it could produce strange results...
    I can give you some new ideas, but as far as I never seen an Icelandic keyboard, and of course don't have one to hack, you have to try some hacking by yourself. I advice you to try (on this order):
        1) May be your keyboard layout is not like a pure QWERTY keyboard. On Brasilian keyboards, we have 49 "main" keys instead of 48. So, I had to include this new key on the layout, generating the 'main_key_scan_abnt_qwerty' and the corresponding 'main_key_vkey_abnt_qwerty' tables, mapping the virtual key VK_OEM_8 to the new key (c cedilla), after ASFGHJKL. You can see these on the current Wine-CVS code. Please notice that the routines that maps keys to vkeys espects to receive some specific VK_OEM_<n> codes for some keys, like '\' (mapped as VK_OEM_7) on X11DRV_InitKeyboard. In fact, the only available VK I found was VK_OEM_8.
        2) on ToUnicode (with my patch applyied), what is the return of ret=XmbLookupString? Please notice that, if the return is zero, it will use the Xlib function  'MultiByteToWideChar' after running a code to map the dead keys. Maybe the KEYBOARD_MapDeadKeysym is not treating your dead_circumflex keycode correctly.
        3) For brazilian keyboards I need only to change XLookupString to XmbLookupString on KEYBOARD_ToUnicode function, in order to use XKB treatment instead of Xlib, but there are some other points where that function is used. Looking at a glance, I haven't found any reason to change, but you can try to understand better the other calls.

    I hope these comments can help you.

Mauro.

Richard Allen wrote:
On Mon, Jan 13, 2003 at 12:07:10PM -0200, Mauro Carvalho Chehab wrote:

I've been studying this quite a bit.  With this patch, I've verifyed that
some of the dead keys work, and some not.   If I change the IS keymap
so Im using dead_circumflex instead of dead_acute, then dead (accented)
characters do work, but obviously not the right ones.
This leads me to think that dead_acute characters are handled differently
by wine than dead_circumflex characters are.

Mauro, do you think I could be on to something here ?



  
Rikki,

   As far as I know, wine keyboard.c uses XFree86 Keymap table to 
compare your key definitions with their internal geometry. After that, 
it maps every scan code to a virtual key. The virtual key  is mapped 
back to a scan code and to an unicode symbol by ToUnicode function. 
During this proccess, it uses part of the original xfree mapping.
   On the original ToUnicode function, it uses the function 
XLookupString (a Xlib function) to map scan code->symbol. This function 
is capable to work with latin characters, but it doesn't implement dead 
keys translation on the current XFree86 implementations. In the past, 
there was some Xlib hacking that changes this behavor (X11 R5 
implementations). The new standard (X11 R6) creates the xkb extensions, 
that provides new functions to treat locale imput data (XmbLookupString, 
Xutf8LookupString and XwcLookupString). What it does is to use XFree86 
locale settings to make this map.
   The dead key treatment is defined on /usr/X11R6/lib/X11/locale.
   The geometry of their keyboards, used by xkb, are on 
/usr/X11R6/lib/X11/xkb (RedHat) or /etc/X11/xkb (Mandrake). The file 
symbols/is, for example, contains some icelandic definitions.
   From your answer, I would try to fix the XFree86 keyboard mapping on 
symbols/is. I guess that this will solve both XFree and Wine (patched).