Re: [PATCH] Make arm_record_test work on big-endian machines

[Simon Marchi <simon dot marchi at polymtl dot ca>]

On 2018-03-12 13:09, Yao Qi wrote:
> Simon Marchi <simon.marchi@ericsson.com> writes:
>
> > What do you think about the patch below?  Functionally, it should be 
> > identical
> > to what you suggested, but I think it's clear to fix extract_arm_insn 
> > instead.
>
> Patch looks good to me.  Thanks.
>
> > +
> > +  insn_record->arm_insn
> > +    = (uint32_t) extract_unsigned_integer (&buf[0], 2, endian);
> > +
> > +  if (insn_size == 4)
> > +    {
> > +      insn_record->arm_insn <<= 16;
> > +      insn_record->arm_insn
> > +	|= (uint32_t) extract_unsigned_integer (&buf[2], 2, endian);
>
> Better to add some comments here.

Oh, actually, does my patch get it wrong for ARM instructions?  For 
them, it seems like in little endian, the four bytes of an instruction 
are reversed (as if it was a 4-byte value).  So that would explain why 
the word swap would only occur under a if (THUMB2_RECORD == 
record_type).  There is no test with ARM instructions in 
arm_record_test... should there be one?

Simon