CBZ to the next instruction on ARM
Mark Shinwell
shinwell@codesourcery.com
Thu Mar 22 17:09:00 GMT 2007
Nick Clifton wrote:
> Hi Paul,
>
>>>> I propose the application of the attached patch that rewrites
>>>> such bogus CBZ instructions to no-ops. This isn't particularly
>>>> elegant, but does make this robust. OK?
>>> The patch is OK, but ... is it necessary to actually generate any
>>> instructions at all ? Wouldn't it be more elegant to just omit
>>> generating an instruction, saving a couple of bytes ?
>>
>> More elegant maybe, but also more surprising. Of course You could
>> argue that the user has written bogus code, so deserve everything they
>> get.
>>
>> It's also significantly more complicated to implement. You have to do
>> insn relaxation rather than simply applying a reloc to a fixed size
>> instruction.
>
> Well I didn't say that it would be easier... :-)
>
> Leaving the patch as-is is OK, but maybe a FIXME could be added so that
> at some later date relaxation could be implemented.
I'll do that then.
Thanks,
Mark
More information about the Binutils
mailing list