Re: [MiNT] typo in keyboard.c

["Helmut Karlowski" <helmut.karlowski@ish.de>]

Am 23.05.2010, 12:45 Uhr, schrieb Alan Hourihane <alanh@fairlite.co.uk>:

> On Sun, 2010-05-23 at 11:21 +0100, Alan Hourihane wrote:
> > On Sun, 2010-05-23 at 09:22 +0200, Helmut Karlowski wrote:
> > > There's a typo in line 160 in keyboard.c:
> > >
> > > static char makes[MAKES_BLEN + 1 * 2];
> > >
> > > should be:
> > >
> > > static char makes[MAKES_BLEN * 2 + 1];
> > >
> > > or even:
> > >
> > > static char makes[MAKES_BLEN * 2 - 1];
> > >
> > > This may cause severe problems. I recommend not to use kernels with  
> > this
> > > error.
> > >
> > > There is a good chance that this also causes eiffel-bug.
> > >
> > > I've committed the fix to my branch.
> >
> > Helmut,
> >
> > I've pulled in this fix and I agree it could cause some unwanted
> > behaviour, but where is the makes[] structure actually used ?
> >
> > It seems to be filled in during ikbd_scan(), but I can't see where
> > makes[] is actually used anywhere else to make use of that data ?
>
> Seems to me as though it was a partial bit of code that never gets used.
> And in fact removing it cleans up a few ifdefs from the MILAN code.
>
> Attached patch.

Currently I don't know completely how this all works, but the subscript of  
makes clearly exceeds 19, and kbd_lock gets overwritten: It has values  
that depend on the typed key.

Does makes[] still have to be 33 bytes long with your patch?

Another issue: what if bioskeys crashes (like it did when calling  
swkbdtbl) and kbd_lock is never reset? The idea to do this in ikbd_scan is  
surely not a perfect solution.

--
Helmut Karlowski