Update for kbd-chooser patch (was: Re: r23706 - in trunk/packages/kbd-choose)
-----BEGIN PGP SIGNED MESSAGE-----
On Tuesday 09 November 2004 04:36, Joey Hess wrote:
> I tried this patch on my ia64, with a usb keyboard, and it does
> something very weird. In expert mode the list of keyboard types, which
> without this patch is "USB keyboard", changes to "e=4384 -- ro"
> The kbd-chooser debug messages from syslog are these:
> mounting usbdevfs to lok for kbd
> Parsing /proc/bus/usb/devices
> Found usbkbd kdb: 0x4a2:0x11
- From your /proc/bus/usb/devices I would expect this to read 'usbhid'.
> non-Apple USB keyboard detected
> keyboard type at: present: false
> keyboard type usb: present: false
These last two lines mean that no keyboard was detected, so you should
have been presented with the option 'No keyboard to configure'.
> I hope this is something simple, like missing a fallback case for !i386
> !powerpc, and not something ugly, like a buffer overflow or stack
> corruption. It seems to work ok aside from the strange choices.
Well, actually there were two problems here.
- - I should have set present to UNKNOWN instead of FALSE; this is closest
to the old situation for other architectures (minor error from me).
- - The existing code that adds a 'no keyboard to configure' option to the
list was broken and would never be added.
I fixed this, but this will lead to that option to be added in a lot of
installations with 2.4 kernels (visible only at medium/low priority).
> FWIW, my /proc/bus/usb/devices:
Thanks for including that.
I've also added the functions to avoid double entries in the arch
selection dialog (as discussed with joeyh yesterday on #d-boot).
I've committed the changes to svn after testing as far as I can, so please
test this new version again.
I'd also appreciate a review of my changes for C syntax.
(bubulle: I did _not_ upload a new mini.iso; please build one yourself)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
-----END PGP SIGNATURE-----