Hi -

> [...] I think I might have located the problem, but am not sure.  If
> I read the code correctly, the Thumb decoder always makes
> entire_insn = base_insn >> 16 ... and if you are
> expecting the next instruction to appear in the upper end of base_insn this
> might make sense.  But if it for some reason appeared in the lower end of
> base_insn, you would throw away the instruction before anything got a chance
> to be executed.  [...]

Yikes.  I guess the thumb part hasn't been tested after it was last
regenerated.  I can fix this, though don't have any arm+thumb
elf binaries handy - do you?

- FChE