can't access array of enums in struct slot

[3b]

(cffi:defcenum (foo :unsigned-int))
(cffi:defcstruct bar
  (f :unsigned-int :count 2)
  (f2 foo :count 2))

(cffi:with-foreign-object (p '(:struct bar))
  (print (cffi:foreign-slot-value p '(:struct bar) 'f)))
=>
#.(SB-SYS:INT-SAP #X7FFFD68CFFE8)

(cffi:with-foreign-object (p '(:struct bar))
  (print (cffi:foreign-slot-value p '(:struct bar) 'f2)))
=>

The value of CFFI::VALUE is #.(SB-SYS:INT-SAP #X7FFFD68CFFF0), which is not of type INTEGER.

Backtrace for: #<SB-THREAD:THREAD "main thread" RUNNING {1003B765A3}>
0: (SB-KERNEL:CHECK-TYPE-ERROR CFFI::VALUE #.(SB-SYS:INT-SAP #X7FFFD68CFFF0) INTEGER NIL)
1: (CFFI::%FOREIGN-ENUM-KEYWORD #<CFFI::FOREIGN-ENUM FOO> #<unavailable argument> :ERRORP T)
2: ((LAMBDA ()))

expected a pointer or sequence of enum keywords

tested on sbcl 1.2.13.2 linux x8664, cffi from dec 2014, and sbcl 1.2.13.48 windows x8664, cffi a bit past v0.15

[Attila Lendvai (attila-lendvai)]

shouldn't it be written something like:

(cffi:with-foreign-object (p '(:struct bar))
              (print (mem-ref p 'foo (foreign-slot-offset '(:struct bar) 'f2))))

i think the given example is a violation of the CFFI API... or is it not?

[Attila Lendvai (attila-lendvai)]

although, on a second thought... even if it's a violation of the API, the original test case should still return with a SAP to the enum array, right?