Re: Modifier mapping in X11

[Adrian Robert]

> Looks like a sort of design change rather than a bug fix.  Though I'm 
> not
> sure whether letting GNUstep depend on X11 explicitly is the right 
> thing
> to do in this particular case,

This is a patch just to part of the X-windows back-end used on 
Unix/Linux systems, not the "GUI" or "Base" libraries that implement 
the OpenStep standard.


>  if GNUstep should do this, I feel your patch
> needs to deal with MappingNotify as well for completeness so that 
> GNUstep
> can properly handle mapping changes via xmodmap or utilities similar 
> to it.

You are right..  I will look into this; I think it's an easy fix.


> Other than facilitating key swappings at this low level, what advantage
> we can take of with this patch?

None.  It only helps you if you happen have two keys mapped to the same 
symbol, like CapsLock and left-control both being mapped to Control_L 
that many people do to improve ergonomics.  It also makes GNUstep apps 
behave more like other X apps without interfering with "GNUstep 
behavior".

Please let me know if you try the patch with any success..


Adrian



On Aug 12, 2004, at 2:32 PM, Kazunobu Kuriyama wrote:

> Adrian Robert wrote:
>
> > > After wondering for a while why emacs-like keybindings were defined 
> > > in my DefaultKeyBindings.dict but not showing up in any apps, I 
> > > realized that GNUstep was apparently not picking up my keymap 
> > > modifications set either with xmodmap or in the XFree86 config file. 
> > >  E.g., I map my caps-lock key to another left-control, but GNUstep 
> > > apps don't pick up control commands, instead inserting characters 
> > > into the text, as in "

".
> > >
> > > I traced this to X11/XGServerEvent.m in 'back'.  It checks modifier 
> > > status by first converting the defaults like "GSFirstControlKey" 
> > > into single X KeyCodes on init, then comparing KeyCodes received 
> > > from the X server with these.  The problem is that X conceptually 
> > > maps modifiers to KeySyms, not KeyCodes. ...
> >
> >
> > > Is there any reason not to make this change?  Will it interact badly 
> > > with any locale / input method code?  If not and no one else wants 
> > > to do it I can take a stab at this and send out a patch for people 
> > > to try..
> >
> >
> > Here is the patch referred to.  Apply in gnustep/core/back using 
> > '-p0'.
> >
> > I've only tested this on my locale="C" machine, so if anyone is using 
> > a different locale and/or especially a different input method their 
> > testing would be most welcome..
> >
> > thanks,
> > Adrian
> >
> > ------------------------