bug#81564: Possible bug when binding dynamically with an existing lexical binding

>   > >   (let ((x 2)) (defvar x) (let ((x 10)) x))
>
> Is it ever reasonable to use defvar inside of let?
> It sure seems weird.

I'm unsure what part specifically you are asking about, but using the
unary `defvar' as a non-toplevel form to declare the symbol as dynamic
locally is definitely legal:

> If INITVALUE is missing, the form marks the variable "special" locally
> (i.e., within the current lexical scope, or the current file, if the
> form is at top-level) [...]

as well as the existence of `dlet'.

It's unclear if the unary `defvar' is allowed when the symbol is
let-bound, but it is explicitly allowed (or accounted for) in the
non-unary case:

> If SYMBOL is let-bound, then this form does not affect the local let
> binding but the toplevel default binding instead, like
> ‘set-toplevel-default-binding‘.

Ideally I would want the behavior of locally special variables to work
as in Common Lisp[1] or at least be capable of doing so.  I don't think
it's likely, since introducing a proper `declare' is too much a hurdle
for just this issue, and the current behavior of `defvar' can't be changed in
a backward-compatible way while supporting such semantics, but I just
want to throw it out there in case there _is_ some way that I am not
thinking of.

[1] https://www.lispworks.com/documentation/HyperSpec/Body/d_specia.htm