cffi-devel June 2009

cffi-devel@common-lisp.net

11 participants
20 discussions

[cffi-devel] New patches: 14-Jun-2009
by Luis Oliveira 15 Jun '09

15 Jun '09

Sun Jun 14 14:16:43 EDT 2009 Luis Oliveira <loliveira(a)common-lisp.net> * clisp: avoid using NIL for null pointers. - Work around FFI:C-POINTER in %MEM-REF, %FOREIGN-FUNCALL and %DEFCALLBACK. - Don't accept NIL in POINTERP, FOREIGN-POINTER or INC-POINTER. - New tests: CALLBACKS.NIL-FOR-NULL, FUNCALL.POINTER-NOT-NIL, POINTER.NULL and FOREIGN-POINTER-TYPE.NIL. M ./src/cffi-clisp.lisp -22 +42 M ./tests/callbacks.lisp +9 M ./tests/funcall.lisp +6 M ./tests/memory.lisp +8 An updated tarball of CFFI's source can be downloaded here: http://common-lisp.net/project/cffi/tarballs/cffi-090614.tar.gz Darcsweb URL: http://common-lisp.net/cgi-bin/darcsweb/darcsweb.cgi?r=cffi-cffi;a=summary

1 0

Re: [cffi-devel] CLISP (pointer-address nil) error
by Luís Oliveira 14 Jun '09

14 Jun '09

On Fri, Jun 12, 2009 at 8:43 AM, Tobias Rautenkranz<tobias(a)rautenkranz.ch> wrote: > Actually I am getting the NIL pointers from cffi:mem-ref: > > CL-USER> (defvar *a* (cffi:foreign-alloc :pointer > :initial-element (cffi-sys:null-pointer))) > *A* > CL-USER> (cffi:mem-ref *a* :pointer) > NIL > > Attached is your patch with the handling for mem-ref added. > > > Getting rid of NIL as null pointer will require additional changes: Thanks. I've pushed a patch based on yours and took care of the additional cases you've identified. Hope that fixes your problems. -- Luís Oliveira http://student.dei.uc.pt/~lmoliv/

1 0

[cffi-devel] *cc* for grovelling on windows+cygwin
by Mirko Vukovic 14 Jun '09

14 Jun '09

Hi, I am running clisp on windows+cygwin., where the gcc compiler is available as `gcc' I am including the skeleton of proposed changes to accommodate this setup: ;;; TODO: figure out a good way to let the user specify where his ;;; compiler is. Can already do that through the CC environment ;;; variable, but that's not very convenient on windows? (defparameter *cc* #+(and clisp windows cygwin) "gcc" #+(and windows (not (and cygwin clisp))) "c:/msys/1.0/bin/gcc.exe" ;; #+windows "c:/msys/1.0/bin/gcc.exe" #-windows "gcc") The reason why I made changes specific to clisp is that I do not know how other lisps announce their cygwin feature. In addition, on windows it is fairly straightforward to set-up an environment variable: In `system properties' (accessible from start/control panel) click on `advanced' tab, and there on the environment variables button. The command-line window needs to be restarted for the changes to take effect (a total reboot should not be necessary) Thanks, Mirko

2 1

[cffi-devel] CFFI ported to ABCL
by Luís Oliveira 14 Jun '09

14 Jun '09

Hello, I have ported CFFI to ABCL using the Java Native Access (JNA) library. This is a preliminary version that will have to be cleaned up and optimized, etc. Nevertheless, at this point it passes all of the relevant CFFI tests except one. Callbacks and shareable vectors are not implemented yet. The former seem to require functionality that is not yet present in ABCL, IIUC. If anyone is interested in playing with this port, I have uploaded a patch to <http://common-lisp.net/~loliveira/patches/cffi-abcl.diff>. It requires trivial-features 0.5. Cheers. -- Luís Oliveira http://student.dei.uc.pt/~lmoliv/

1 0

[cffi-devel] CLISP (pointer-address nil) error
by Tobias Rautenkranz 11 Jun '09

11 Jun '09

Hi CLISP returns NIL as null pointer but: (cffi:pointer-address nil) gives an error. Thanks for fixing this. (see the patch below) diff -rN -u old-cffi/src/cffi-clisp.lisp new-cffi/src/cffi-clisp.lisp --- old-cffi/src/cffi-clisp.lisp 2009-06-11 21:43:09.816010652 +0200 +++ new-cffi/src/cffi-clisp.lisp 2009-06-11 21:43:09.860018215 +0200 @@ -140,7 +140,9 @@ (defun pointer-address (ptr) "Return the address pointed to by PTR." - (ffi:foreign-address-unsigned ptr)) + (if ptr + (ffi:foreign-address-unsigned ptr) + 0)) ;;;# Foreign Memory Allocation

2 1

[cffi-devel] New patches: 8-Jun-2009
by Luis Oliveira 09 Jun '09

09 Jun '09

Mon Jun 8 10:29:44 EDT 2009 Stelian Ionescu <sionescu(a)common-lisp.net> * Groveler: add support for float constants. The CONSTANT groveler form now takes an additional keyword argument TYPE, which defaults to INTEGER. In order to grovel floats, DOUBLE-FLOAT must be used. M ./grovel/grovel.lisp -5 +18 An updated tarball of CFFI's source can be downloaded here: http://common-lisp.net/project/cffi/tarballs/cffi-090608.tar.gz Darcsweb URL: http://common-lisp.net/cgi-bin/darcsweb/darcsweb.cgi?r=cffi-cffi;a=summary

1 0

Re: [cffi-devel] stdcall callbacks on ClozureCL/win32
by Luís Oliveira 08 Jun '09

08 Jun '09

On Sat, Jun 6, 2009 at 12:40 PM, Greg Santucci<thecodewitch(a)gmail.com> wrote: > When I run (rt:do-test 'cffi-tests::callbacks.stdcall.1), I get this > printed: > > CFFI-TESTS::CALLBACKS.STDCALL.1 > > and nothing more. That's good, it means the test passed. > When I try: > > (asdf:oos 'asdf:load-op :cffi-tests) > (asdf:oos 'asdf:test-op :cffi-tests) > > Clozure crashes to its kernel debugger. Can you get a backtrace? Hopefully one of the tests is triggering the same bug you're experiencing with GLUT which I suspect is unrelated to stdcall. -- Luís Oliveira http://student.dei.uc.pt/~lmoliv/

2 6

[cffi-devel] New patches: 4-Jun-2009
by Luis Oliveira 05 Jun '09

05 Jun '09

Thu Jun 4 09:05:25 EDT 2009 Luis Oliveira <loliveira(a)common-lisp.net> * cffi-tests.asd: use trivial-features. - Using #-windows instead of #-(or win32 mswindows) which didn't work on Clozure CL. M ./cffi-tests.asd -1 +4 Thu Jun 4 08:42:41 EDT 2009 Luis Oliveira <loliveira(a)common-lisp.net> * Add stdcall support for callbacks in Clozure CL. M ./src/cffi-openmcl.lisp -2 +3 An updated tarball of CFFI's source can be downloaded here: http://common-lisp.net/project/cffi/tarballs/cffi-090604.tar.gz Darcsweb URL: http://common-lisp.net/cgi-bin/darcsweb/darcsweb.cgi?r=cffi-cffi;a=summary

1 0

[cffi-devel] stdcall callbacks on ClozureCL/win32
by Gary Byers 04 Jun '09

04 Jun '09

Hi. In the thread starting in: <http://common-lisp.net/pipermail/cl-opengl-devel/2009-June/000585.html> someone described problems running some CL-OPENGL GLUT examples in Clozure CL on Win32. I'm not sure if this is the only problem, but it seems that the most functions defined in the FreeGLUT library use the 'stdcall' calling convention (a "Pascal"-like calling sequence where the callee is responsible for discarding args from the stack, sometimes used on Win32 for no obvious reason besides backward compatibility ...) and it seems reasonable to assume that it expects user-defined functions that it calls to follow this convention as well. Because of implementation arcana, CCL doesn't really care whether a foreign function that it calls follows 'stdcall' or 'cdecl' conventions. The caller of a callback function (defined via CCL:DEFCALLBACK or the CFFI equivalent) may very well care, and the callback machinery has to be told which conventions to follow. A callback function that accepts two 'int' arguments and returns their sum as an 'int' would be defined in CCL as: (ccl:defcallback add-two-ints (:int x :int y :int) ;; The arglist declares X and Y to be C ints and that the function returns ;; a value of type 'int'. (+ x y)) That makes the value of ADD-TWO-INTS a foreign pointer to foreign-callable code that ultimately just adds a couple of signed 32-bit integers. By default, the glue code that handles the transition from C to lisp and back again when this function is called from foreign code assumes that pretty standard C calling conventions are being followed (all args are passed on the stack and left there when the callee returns, with a possible exception involving the invisible argument used to effect structure return.) If we wanted to define a callback that followed 'stdcall' conventions, we can do so by prepending the keyword :DISCARD-STACK-ARGS to the callback's argument list: (ccl:defcallback stdcall-add-two-ints (:discard-stack-args :int x :int y :int) ;; note that there's just a :DISCARD-STACK-ARGS keyword at the front of the ;; arglist, not a key/value pair (+ x y)) I'm fairly sure that the runtime effects of :DISCARD-STACK-ARGS will only occur on Win32 (e.g., that it's a no-op on other platforms). I'd assume that anyone who's: a) more familiar with CFFI than I am and b) less lazy than I am can add support for stdcall-style callbacks to CFFI's OpenMCL/CCL-backend code, and doing so might help to get things like CL-OPENGL working on CCL/win32. If anyone has any questions abou the CCL side of this, I'd be glad to try to answer them. It looks like CFFI is generally familiar with this cdecl/stdcall stuff and has the infrastructure in place to do the right thing, but the CCL backend just doesn't implement it.

2 1

[cffi-devel] Grovel float constants
by Liam Healy 04 Jun '09

04 Jun '09

I am trying to use cffi-grovel on GSL's physical constants. For example, in http://git.savannah.gnu.org/cgit/gsl.git/tree/const/gsl_const_mksa.h, these are defined: #define GSL_CONST_MKSA_SPEED_OF_LIGHT (2.99792458e8) /* m / s */ #define GSL_CONST_MKSA_GRAVITATIONAL_CONSTANT (6.673e-11) /* m^3 / kg s^2 */ When I try to define these using CFFI-grovel (constant (+mksa-speed-of-light+ "GSL_CONST_MKSA_SPEED_OF_LIGHT")) I end up with an integer, because the C definition is cast to an integer prior to generating the constant. Is there a way to get around this, so that I can grovel float constants? Thanks, Liam

2 4

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

cffi-devel June 2009