Re: [PATCH v4] Add support for creating new types from the Python API

[Matheus Branco Borella <dark.ryu.550@gmail.com>]

> How can that work?  AFAIU, most architectures only allow pointers of
> certain sizes, some allow pointers of just one size.  E.g., what
> happens if I create a 16-bit pointer on a 64-bit target?

My intention is primarily to make it possible to construct such types,
assuming the given configuration is valid. In this case it would be the
consumer of the API who would be responsible for guaranteeing what they
are doing is valid.

Regardless, as far as I could gather, GDB doesn't really seem to care
if values whose types have TYPE_CODE_PTR have sizes that are valid in
the target architecture. `unsigned_pointer_to_address`, as well as all
the `*_pointer_to_address` functions I could find by grepping for uses
of `set_gdbarch_pointer_to_address` in the code are perfectly fine just
reading the data in the pointers as if they were type->length()-sized
integers. And mostly the same goes for their `*_address_to_pointer`
counterparts, except for large values being clipped (including extra
function pointer information). But I believe that behavior should be
fairly unsurprising if you're creating a pointer with the wrong size.

So, to answer your question, AFAICT it would just treat it as an
address stored in an (u)int16_t.