Re: [PATCH 0/2] Fix ODR violations in regformat target descriptions

[Keith Seitz <keiths@redhat.com>]

On 6/18/26 10:10 AM, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:
> 
> Simon> Other than the little nit I sent on patch 1, this series LGTM (although
> Simon> I still don't fully understand why regdat exists and how it works).
> 
> I was curious about this as well.  Like, how hard is it to remove this?
> Since IIUC it's an obsolete way of writing a target description.

 From what I understand while hacking at this, you are correct. It is a
very outdated way of writing a target description. The modern way is via
XML/features.

I asked Claude to explain what would it would take to get rid of this
and update everything. Unless it is completely fabricating its response
(and I have not read anything to suggest such), it is a non-trivial
amount of work.

The short of it is, there are six "legacy" targets with no XML backing
whatsoever. These targets include padding registers, integer-only
registers (missing FP), and other peculiarities which may be non-trivial
to implement completely and/or correctly. [Or not -- I've not dug *that*
deeply into it.]

There are an additional 42 targets which are XML-backed that would also
require some conversion. Testing all of these will be a nightmare.

If you'd like, I can file a bug and add the complete Claude response.
Just let me know.

In the meantime, I've pushed these patches. Thank you to you and Simon
for your reviews.

Keith