[PATCH] xkb: Do not use base group as an array index.

[Andreas Wettstein]

> is there some sort of test-case that triggers this issue reliable?

I found it when I tried:

key <FK07> {
    type= "ONE_LEVEL",
    symbols[Group1]= [              NoSymbol ],
    actions[Group1]= [ LatchGroup(group=-1, clearLocks) ]
};

and then hit F7.  Using SetGroup(group=-1) should "work" equally well.

> also, why do we ignore locked groups here? It'd be great to have some
> explanation, this is code that hardly anyone knows inside-out, so having
> this in the commit message would help a lot.

Actually, at first I fixed the crash using the effective group.  But
then I found in section 2.3.1 of the X Keyboard Extension Protocol
Specification:

  If the IgnoreGroupLock control is set, the locked state of the
  keyboard group is not considered when activating passive grabs.

I believe that this is what this piece of code is about, and I think my
choice is as close as possible to this description.

And what is the idea behind the specification?  Probably, the idea is
that users switching between two distinct two-level layouts (such as us
and ru) using a lock can have that lock partially ignored, while users
using a single three level layout implemented using two groups (for core
protocol compatibility) will not get Mode_switch ignored, which makes
sense, as it actually selects symbols of the same layout.

Regards,

Andreas