On 2025-10-20 17:08 +0000, Sergei Egorov wrote:
> > I second (third?) Daphne & Marc here. If the matching order is
> > unspecified, then ban variable references that depend on order.
>
> They are banned already, that was the intent from day 0 (if you are
> talking about pattern var references).

Understood.  But in that case, it's very confusing that the sample
implementation allows such references.  I know that providing a
"non-portable extension" is one possibility for cases of UB, but an
extension that creates surprising intra-pattern bindings seems like
a bad idea, in principle.  I would much prefer that the implementation
signal an error.

--
Wolfgang Corcoran-Mathe  <xxxxxx@sigwinch.xyz>