[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Problem with FFI in CLISP-1997-08-07



Hi,

using the newest CLISP (1997-08-07) on my Linux Box (Linux 2.0.30,
PPro200, 64MB, libc.so.5.4.17) I get severe problems, with FFI CALL-INs.
The following code is a minimal example to reproduce the error:

==== CUT HERE ================================================================
;;; --------------------------------------------------------------------------
;;; File: ltffi.lsp
;;; --------------------------------------------------------------------------

(in-package "TEST-FFI")

(defpackage TEST-FFI
  (:nicknames TFFI)
  (:use       COMMON-LISP FFI)
  (:export    DO-TEST))

(defun FFI::CALL-IN-FUN (msg)
  (format NIL "CALL-IN RECEIVED: ~S.~%" msg))

(def-c-call-in  FFI::CALL-IN-FUN
  (:name        "lisp_call_in")
  (:arguments   (msg c-string :in :malloc-free))
  (:return-type c-string :malloc-free))

(def-c-call-out DO-TEST
  (:name        "tffi_test")
  (:arguments   (msg c-string)))

==== CUT HERE ================================================================
/*----------------------------------------------------------------------------
 * File: tffi.c
 *----------------------------------------------------------------------------
 */
void 
tffi_test(char *message)
{
    extern char *lisp_call_in(char *msg);

    printf("lisp_call_in(%s) -> %s\n", message, lisp_call_in(message));
}
==== CUT HERE ================================================================

Here is, what I do to compile and bind a new clisp:

% make
/src/lisp/clisp/mem/ffi/lisp.run -q -M /src/lisp/clisp/mem/ffi/lispinit.mem -c ltffi.lsp

Compiling file /src/lisp/clisp/test-ffi/ltffi.lsp ...

Compilation of file /src/lisp/clisp/test-ffi/ltffi.lsp is finished.
0 errors, 0 warnings
gcc -g -Wall   -I/src/lisp/clisp/mem/ffi   -c ltffi.c -o ltffi.o
In file included from ltffi.c:1:
/src/lisp/clisp/mem/ffi/clisp.h:2744: warning: register used for two global register variables
gcc -g -Wall   -I/src/lisp/clisp/mem/ffi   -c tffi.c -o tffi.o
tffi.c: In function `tffi_test':
tffi.c:6: warning: implicit declaration of function `printf'
rm -f modules.h
for i in  ltffi; do \
  echo "MODULE($i)" >> modules.h; \
        done
gcc -g -Wall   -I/src/lisp/clisp/mem/ffi -I. -c /src/lisp/clisp/mem/ffi/modules.c -o modules.o
In file included from modules.d:11:
/src/lisp/clisp/mem/ffi/clisp.h:2744: warning: register used for two global register variables
gcc -o lisp.run modules.o /src/lisp/clisp/mem/ffi/lisp.a ltffi.o tffi.o \
        -L/src/lisp/clisp/mem/ffi -lavcall -lvacall -ltrampoline -lintl -ltermcap -lreadline
lisp.run -q -M /src/lisp/clisp/mem/ffi/lispinit.mem -x '(dolist (f (list "./ltffi.fas")) (lisp:gc) (load f)) (lisp:gc) (lisp:saveinitmem) (user::bye)' < /dev/null
;; Loading file ./ltffi.fas ...
;; Loading of file ./ltffi.fas is finished.
NIL
524288
729604 ;
524288
%

Using CLISP-1997-05-03 I get the expected result:

> (lisp-implementation-version)
"1997-05-03 (May 1997)"
> (tffi:do-test "CLISP")
lisp_call_in(CLISP) -> CALL-IN RECEIVED: "CLISP".


> (Bye)
Bye.

The new clisp however gets a SIGSEGV and dies.
Is this a bug, or is there something wrong with my setup or do I
misuse the ffi? 
I have tried to disable generational GC, but this did not change
anything.

Another question:
   I am writing some kind of daemon using clisp. It uses the ffi a
   lot (mySQL interface, tcpip interface etc.)  and it would be very
   helpful, if there was a way to deal with SIGSEGVs within clisp
   (do a restart e.g.). Is there a "canonical" way to do this?


Any help appreciated

--Matthias

------------------------------------------------------------------------------
Matthias Lindner
Wilhelm-Leuschner-Str.46
64293 Darmstadt
Germany
TEL: +49 6151 22071
NET: matthias@intellektik.informatik.th-darmstadt.de
WWW: http://aida.intellektik.informatik.th-darmstadt.de/~matthias/
------------------------------------------------------------------------------