Re: [PATCH RESEND v2] Aarch64: Allow explicit size specifier for predicate operand of {sq, uq, }{incp, decp}

[Richard Sandiford <richard.sandiford@arm.com>]

Jan Beulich <jbeulich@suse.com> writes:
> On 12.09.2022 10:17, Richard Sandiford via Binutils wrote:
>> It's extremely unlikely that the old syntax would be repurposed to mean
>> something else.  Therefore, Richard and I thought that it might be better
>> to continue to accept the old syntax alongside the new syntax in the GNU
>> tools, as a GNU extension to the official assembler spec, without the
>> deprecation message.  Sorry to raise this after so long, and after you'd
>> already done the work.
>
> Isn't it the case already that the old syntax will continue to be accepted,
> by there being new insn entries added without removing the original ones?

Yes, for now.  But the point of deprecation is that it anticipates
removing the old syntax in future.  The suggestion here was instead
to keep the old syntax alongside the new one, without any intention
of removing it.

> IOW is it perhaps that you're asking for the old syntax to remain accepted
> _silently_ (and don't claim this behavior will be removed)? If so, I'd like
> to suggest to perhaps allow for diagnostics, but require a command line
> option to turn them on.

Richard suggested having a “strict” flag that (IIUC) would make gas
stick more rigidly to the ISA definition.  But I think that would be
separate work and would affect quite a few more instructions.  E.g. it
should probably insist on curly brackets around the Z registers in SVE
LD1 and ST1 instructions.  Perhaps it should even insist on # for
immediate operands.

I don't think it's worth having an option specifically for these
instructions in isolation.  And if we did add the strict flag,
I think we should try to make the first implementation relatively
complete, so that its behaviour is somewhat predictable between releases.

Thanks,
Richard