Frans Pop wrote:
On Thursday 30 November 2006 17:07, Attilio Fiandrotti wrote:Today, after many unlucky experiments, i tried disabling the "keyboard" module in directfbrc, hence having linux_input manage the keyboard, as DOK and suggested some times ago and Davide reminded me yesterday. Guess guess, now keymap can be correctly switched multiple times between english and italian, and i also verified thatOK. This does indeed seem to fix the "multiple switching problem". I have tested both in vmware and with an usb keyboard on a normal system and both worked OK. AltGr key combinations seem to work fine.However, it does _not_ fix the problems with dead keys described in #394871 (tested with French and (customized) Dutch keymaps).
sure, that's another bug and has to be fixed differently: i successfully backported from DirectFB CVS the part of code needed to manage dead keys but no luck yet, i'll try again tomorrow.
Note that I'm still somewhat unhappy with "solutions" like this. Upstream directfb developers should IMHO still fix the problems we are seeing when the keyboard module is not disabled (or they should disable the keyboard module by default).However, it does seem to be a real improvement, so we should probably implement this for Etch.
I believe that for the long term we should aim at managing all mices, touchpads and keyboards on all architectures via linux_input only, which relies upon the evdev input interface. Getting there will require a lot of upstream work for post-etch : ATM linux_input is not usable on some IBooks, where we had to force-disable it.
As we *do not* want keyboard module to be disabled on PPC too i'll prepare a script for rootskel-gtk that takes care of disabling that module on i386 and AMD64 only.
cheers Attilio