DEFCTYPE alias for struct skips type translations

[Jan Igger]

This is more a question than a bug probably, but I don't know where else to ask, so why does this work as follows?

```
(cffi:defctype uint8 :unsigned-char)

(cffi:defcstruct (anon-struct-39 :size 16)
  (data (:array uint8 16) :offset 0))

(cffi:defctype joystick-guid (:struct anon-struct-39))

(cffi:convert-from-foreign
 (foreign-alloc 'joystick-guid)
 '(:struct anon-struct-39))
;; => (DATA #(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0))

(cffi:convert-from-foreign
 (foreign-alloc 'joystick-guid)
 'guid)
;; => #.(SB-SYS:INT-SAP #X7FC1E8001950)
```

Why can't I just use 'guid in convert-from-foreign?

[Jan Igger (le-perv)]

Sorry, here's a cleaned-up version:

(cffi:defctype uint8 :unsigned-char)

(cffi:defcstruct (anon-struct :size 16)
  (data (:array uint8 16) :offset 0))

(cffi:defctype guid (:struct anon-struct))

(cffi:convert-from-foreign
 (foreign-alloc 'guid)
 '(:struct anon-struct))
;; => (DATA #(0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0))

(cffi:convert-from-foreign
 (foreign-alloc 'guid)
 'guid)
;; => #.(SB-SYS:INT-SAP #X7FC1E8001950)

[Luís Oliveira (luismbo)]

Someone tried to fix this before but we had to revert it because it caused other problems: https://github.com/cffi/cffi/pull/152

There are some inconsistencies in the struct-as-values code is what I recall. Definitely a bug.

[Jan Igger (le-perv)]

Hm, alright, thanks for the fast response!

Also, in SDL, with the definitions as in the first post held in mind, there's a function:

SDL_JoystickGUID SDL_JoystickGetGUID(SDL_Joystick *joystick)
{
    if (!SDL_PrivateJoystickValid(joystick)) {
        SDL_JoystickGUID emptyGUID;
        SDL_zero(emptyGUID);
        return emptyGUID;
    }
    return joystick->guid;
}

where SDL_zero is:

#define SDL_zero(x) SDL_memset(&(x), 0, sizeof((x)))

and with the code:

(cffi:defcfun ("SDL_JoystickGetGUID" joystick-get-guid)
    joystick-guid
  (joystick (:pointer joystick)))

The plist that

(joystick-get-guid (cffi:null-pointer))

returns, for some reason, consistenly contains non-zero values:

(DATA #(0 0 0 0 0 0 0 0 195 28 0 0 0 0 0 0))

What's interesting, memset'ing a manually allocated joystick-guid and doing convert-from-foreign on it yields the correct result with all zeroes.

Should I file a seperate issue for this?

[Luís Oliveira (luismbo)]

Yes, that seems like a separate issue. If you could reduce the test case to something that doesn't depend on SDL, that'd be nice. BTW, the cffi-devel mailing list is a good place (if out of fashion) to have preliminary discussions like this.

[Jan Igger (le-perv)]

OK, will do!

Concerning the automatic conversion of structs to plists on function return, is this a part of stable API at this point, or can this still change in the future? If you want, I can ask this on the mailing list instead.

[Luís Oliveira (luismbo)]

Yes, but you can customize how structs are returned. https://common-lisp.net/project/cffi/manual/html_node/Foreign-Structure-Types.html

[Luís Oliveira (luismbo)]

Jan, https://github.com/cffi/cffi/pull/181 may fix this. Can you take it for a spin?

[Jan Igger (le-perv)]

Works as expected now! Thank you!