Re: ARM: Improved IT disassembly

[Zack Weinberg <zack at codesourcery dot com>]

Richard Earnshaw <Richard.Earnshaw@arm.com> writes:

>> I'm honestly at a loss for any other viable approach.  The instruction
>> encoding doesn't lend itself to scanning backward, and in any case, in
>> print_insn I have no information to say where a backward scan should
>> stop.
>
> There is a reasonable bound (any code label must be on an instruction
> boundary) and scanning back should converge rapidly.  I'll send you a
> private mail with the details, but it will have to wait until tomorrow
> now.

To be slightly clearer, when I said "I have no information to say
where a backward scan should stop", I meant "print_insn does not know
the beginning of the memory region containing the instruction."  If I
blindly call info->read_memory_func (pc - 2), will it not crash in the
case where pc is at the very beginning of the segment?

zw