[Ecls-list] ECL & static symbol references with CFFI

[Sylvain Ageneau]

Doing the following change seems to fix the issue:

diff --git a/src/cffi-ecl.lisp b/src/cffi-ecl.lisp
index 9f9500b..a31d069 100644
--- a/src/cffi-ecl.lisp
+++ b/src/cffi-ecl.lisp
@@ -290,9 +290,10 @@ WITH-POINTER-TO-VECTOR-DATA."
                 ;; On AMD64, the following code only works with the extra
                 ;; argument ",...". If this is not present, functions
                 ;; like sprintf do not work
-                (format s "extern ~A ~A(~@[~{~A~^, ~}~]);
-@(return) = ~A(~A);"
+                (format s "{extern ~A ~A(~@[~{~A~^, ~}~]);
+~A~A(~A);}"
                         (ecl-type->c-type return-type) pointer types
+                        (if (eq return-type :void) "" "@(return) =")
                         pointer
                         (subseq +ecl-inline-codes+ 0
                                 (max 0 (1- (* (length values) 3)))))))


On Nov 22, 2012, at 3:05 PM, Sylvain Ageneau <ageneau at gmail.com> wrote:

> Sorry that should have been:
> (cffi:foreign-funcall "srand" :unsigned-int 55 :void)
> 
> But the issue happens the same.
> 
> On Nov 22, 2012, at 2:41 PM, Sylvain Ageneau <ageneau at gmail.com> wrote:
> 
>> I think there's an issue with void functions, for example:
>> 
>> (cffi:defcfun ("srand" c-srand) :void '(:unsigned-int))
>> 
>> (defun test-srand ()
>>   "invokes srand function"
>>   (cffi:foreign-funcall "srand"))
>> 
>> (test-srand)
>> 
>> 
>> ;;;
>> ;;; Compiling cffi-test.lisp.
>> ;;; OPTIMIZE levels: Safety=2, Space=0, Speed=3, Debug=0
>> ;;;
>> ;;; Compiling (CFFI:DEFCFUN ("srand" C-SRAND) ...).
>> ;;; Compiling (DEFUN TEST-SRAND ...).
>> ;;; End of Pass 1.
>> ;;; Emitting code for C-SRAND.
>> ;;; Error:
>> ;;;   in file cffi-test.lisp, position 415
>> ;;;   at (DEFCFUN (srand C-SRAND) ...)
>> ;;;   * Used @(RETURN 0) in a C-INLINE form with 0 output values
>> ;;; Loading "/Users/sylvain/src/LISP/cffi_ecl/cffi-test.fas"
>> 
>> 
>> 
>> On Nov 22, 2012, at 11:53 AM, Sylvain Ageneau <ageneau at gmail.com> wrote:
>> 
>>> Great!
>>> 
>>> I'll try that and report back. Many thanks.
>>> 
>>> On Nov 22, 2012, at 11:16 AM, Juan Jose Garcia-Ripoll <juanjose.garciaripoll at gmail.com> wrote:
>>> 
>>>> I am attaching a patch for CFFI that allows ECL to build code that does not use dlopen() but rather uses the function symbols as they are know to the C compiler. This is best explained in the code comments, which I reproduce below, and in the example.
>>>> 
>>>> The patched CFFI is available at my github account
>>>> https://github.com/juanjosegarciaripoll/cffi
>>>> 
>>>> This is how the example is used
>>>> 
>>>> bash-3.2$ ecl -norc
>>>> ECL (Embeddable Common-Lisp) 12.7.1 (git:e95fa190ccab10a5dcca791658d7d3ff24a77240)
>>>> Copyright (C) 1984 Taiichi Yuasa and Masami Hagiya
>>>> Copyright (C) 1993 Giuseppe Attardi
>>>> Copyright (C) 2000 Juan J. Garcia-Ripoll
>>>> ECL is free software, and you are welcome to redistribute it
>>>> under certain conditions; see file 'Copyright' for details.
>>>> Type :h for Help.  
>>>> Top level in: #<process TOP-LEVEL>.
>>>> > (load "build")
>>>> [...]
>>>> To load "cffi":
>>>>   Load 1 ASDF system:
>>>>     cffi
>>>> ; Loading "cffi"
>>>> 
>>>> ;;;
>>>> ;;; Compiling cffi-test.lisp.
>>>> ;;; OPTIMIZE levels: Safety=2, Space=0, Speed=3, Debug=0
>>>> ;;;
>>>> ;;; Compiling (CFFI:DEFCFUN ("sin" C-SINE) ...).
>>>> ;;; Compiling (DEFUN FF-C-SINE ...).
>>>> ;;; End of Pass 1.
>>>> ;;; Emitting code for C-SINE.
>>>> ;;; Emitting code for FF-C-SINE.
>>>> ;;; Finished compiling cffi-test.lisp.
>>>> ;;;
>>>> ;;; Loading "/Users/jjgarcia/build/cffi-static/cffi-test.fas"
>>>> 
>>>>  sin(pi) = 1.22464679914735320720000000000000d-16
>>>> #P"/Users/jjgarcia/build/cffi-static/build.lisp"
>>>> 
>>>> This is the self-explanatory comment
>>>> 
>>>> ;;;
>>>> ;;; ECL allows many ways of calling a foreign function, and also many
>>>> ;;; ways of finding the pointer associated to a function name. They
>>>> ;;; depend on whether the FFI relies on libffi or on the C/C++ compiler,
>>>> ;;; and whether they use the shared library loader to locate symbols
>>>> ;;; or they are linked by the linker.
>>>> ;;;
>>>> ;;;  :DFFI
>>>> ;;;
>>>> ;;;  ECL uses libffi to call foreign functions. The only way to find out
>>>> ;;;  foreign symbols is by loading shared libraries and using dlopen()
>>>> ;;;  or similar.
>>>> ;;;
>>>> ;;;  :DLOPEN
>>>> ;;;
>>>> ;;;  ECL compiles FFI code as C/C++ statements. The names are resolved
>>>> ;;;  at run time by the shared library loader every time the function
>>>> ;;;  is called
>>>> ;;;
>>>> ;;;  :C/C++
>>>> ;;;
>>>> ;;;  ECL compiles FFI code as C/C++ statements, but the name resolution
>>>> ;;;  happens at link time. In this case you have to tell the ECL
>>>> ;;;  compiler which are the right ld-flags (c:*ld-flags*) to link in
>>>> ;;;  the library.
>>>> ;;;
>>>> (defvar *cffi-ecl-method*
>>>>   #+dffi :dffi
>>>>   #+(and dlopen (not dffi)) :dlopen
>>>>   #-(or dffi dlopen) :c/c++
>>>>   "The type of code that CFFI generates for ECL: :DFFI when using the
>>>> dynamical foreign function interface; :DLOPEN when using C code and
>>>> dynamical references to symbols; :C/C++ for C/C++ code with static
>>>> references to symbols.")
>>>> 
>>>> 
>>>> -- 
>>>> Instituto de Física Fundamental, CSIC
>>>> c/ Serrano, 113b, Madrid 28006 (Spain) 
>>>> http://juanjose.garciaripoll.googlepages.com
>>>> <build.lisp><cffi-test.lisp><0001-Allow-the-ECL-backend-to-support-different-combinati.patch>
>>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/ecl-devel/attachments/20121122/0ff86e14/attachment.html>