[PATCH v2] RISC-V: Change -march parsing.

[Palmer Dabbelt via binutils]

On Wed, 08 Jan 2020 22:20:36 PST (-0800), Kito Cheng wrote:
> Hi Palmer:
>
> Ping, do you mind accept this patch and then we'll send follow up
> patch for Zifence and Zicsr?

Sorry, I must have missed the original email

>
> On Mon, Dec 23, 2019 at 3:21 PM Kito Cheng <kito.cheng@gmail.com> wrote:
>>
>> Hi Maxim:
>>
>> LGTM, just need Palmer or Jim to approve :)
>>
>> Hi Palmer:
>>
>> >> +static bfd_boolean
>> >> +riscv_ext_x_valid_p (const char *arg)
>> >> +{
>> >> +  if (!strcasecmp (arg, "x"))
>> >> +    return FALSE;
>> >> +
>> >> +  return TRUE;
>> >
>> > This allows any X extension through but then goes ahead to not provide any
>> > support for those extensions.  I can understand providing some scaffolding to
>> > users to turn on their custom extensions, but we should fail quickly on any
>> > extensions that aren't supported.
>>
>> We bypass any X extension now, and one concern is we can't testing any
>> X parsing/combine logic if we reject that.

I still don't think it's a good idea to allow any X extension through.  If you
really want to test X extension handling then we should define some X extension
that's unlikely to have a name conflict and have that extension do something --
maybe "Xbinutils_example_extension", which only has some sort of
"Xbinutils_example_instruction" instruction in it.  Without that then you're
not really testing X extension parsing, as there won't be any behavior change
as a result of passing in the X extension so you can't tell if it's actually
enabled.