CFFI callbacks don't unbox integer values

[stassats]

It very well may be a cffi issue.

Compile the following C code
with gcc -shared test.c -o test.so

#ifdef WIN32
#define DLLEXPORT __declspec(dllexport)
#else
#define DLLEXPORT
#endif

DLLEXPORT
short test(short (*f)(short, short))
{
 return f(23, 24);
}

Then

(cffi:load-foreign-library "/tmp/test.so")

(cffi:defcfun test :short
  (function :pointer))

(cffi:defcallback (test-callback)
    :short ((a :short) (b :short))
  (+ a b))

And finally

(test (cffi:get-callback 'test-callback)) =>
The value #<java.lang.Short 23 {5EE76547}> is not of type NUMBER.

That's on 64-bit x86 linux, with OpenJDK 1.6.0_24 and jna downloaded by maven.

[Mark Evenson]

Need some basic macrology--and a reasonable abstraction of type-- in [ABCL's cffi-sys:convert-args-to-lisp-values][cffi-abcl]

[cffi-abcl]: ​https://github.com/easye/cffi/commit/c630aaf3b41501d3d2ac60b94461f96627fce627

No longer critical, as CL+SSL at least sort of loads. Callbacks still fail (sometimes) until the types are properly reasoned about.

[Mark Evenson]

Problem is with upstream cffi abcl implementation, so does not block abcl-1.1.

[Mark Evenson]

pointerp and null-pointer-p don't work on non-pointers

[stassats]

pointerp and null-pointer-p are not related.
The problem now with the return value from the callback.

JNA: Callback $Proxy2@7dc6a657 threw the following exception:
java.lang.ClassCastException: java.lang.Integer cannot be cast to java.lang.Short

[Evenson Not Org]

No longer blocks hunchentoot, but CFFI stills fails when actually trying to use CL+SSL.

[Mark Evenson]

Ticket retargeted after milestone closed

[Mark Evenson]

Milestone renamed

[Mark Evenson]

Ticket retargeted after milestone closed