- cffi-devel - mailman3.common-lisp.net

[cffi-devel] EPoll Interface
by Volkan YAZICI 11 Sep '09

11 Sep '09

Hi, I'm working on an epoll interface, but having trouble with nested structs. Below code snippet is just a minor example where things explode. Does anybody have an idea about what I might be missing? Regards. (defmacro prog1-let ((var val) &body body) `(let ((,var ,val)) ,@body ,var)) (defmacro maybe-error (test) (let ((errno (gensym))) `(when ,test (let ((,errno *errno*)) (error (foreign-funcall "strerror" :int ,errno :string)))))) (defcvar "errno" :int) ; typedef union epoll_data { ; void *ptr; ; int fd; ; __uint32_t u32; ; __uint64_t u64; ; } epoll_data_t; (defcstruct data (ptr :pointer) (fd :int) (u32 :uint32) (u64 :uint64)) ; struct epoll_event { ; __uint32_t events; /* Epoll events */ ; epoll_data_t data; /* User data variable */ ; }; (defcstruct event (events :uint32) (data data)) ; int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event); (defcfun "epoll_ctl" :int (epfd :int) (op :int) (fd :int) (event (:pointer event))) (defun control (epfd op fd events) (with-foreign-object (event-ptr 'event) ;; Set `EVENTS' slot of the `EVENTS'. (setf (foreign-slot-value event-ptr 'event 'events) events) ;; Empty `DATA' slot of the `EVENTS', and set `FD' slot of the `DATA'. (let ((data-ptr (foreign-slot-pointer event-ptr 'event 'data))) (setf (foreign-slot-value data-ptr 'data 'ptr) 0 (foreign-slot-value data-ptr 'data 'u32) 0 (foreign-slot-value data-ptr 'data 'u64) 0 (foreign-slot-value data-ptr 'data 'fd) fd)) (prog1-let (res (foreign-funcall "epoll_ctl" :int epfd :int op :int fd :pointer event-ptr :int)) (maybe-error (minusp res))))) ; in: DEFUN CONTROL ; (SETF (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::PTR) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::U32) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::U64) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::FD) ; EPOLL::FD) ; --> PROGN SETF LET* MULTIPLE-VALUE-BIND LET PROGN CFFI::FOREIGN-SLOT-SET ; --> SETF LET* MULTIPLE-VALUE-BIND LET PROGN CFFI::MEM-SET ; --> CFFI-SYS:%MEM-SET LET SETF SB-KERNEL:%SET-SAP-REF-32 ; ==> ; EPOLL::DATA-PTR ; ; note: deleting unreachable code ; (EPOLL::PROG1-LET ; (EPOLL::RES ; (CFFI:FOREIGN-FUNCALL "epoll_ctl" :INT EPOLL::EPFD :INT EPOLL::OP :INT ; EPOLL::FD :POINTER EPOLL::EVENT-PTR :INT)) ; (EPOLL::MAYBE-ERROR (< EPOLL::RES 0))) ; --> LET ; ==> ; EPOLL::RES ; ; note: deleting unreachable code ; (SETF (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::PTR) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::U32) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::U64) ; 0 ; (CFFI:FOREIGN-SLOT-VALUE EPOLL::DATA-PTR 'EPOLL::DATA 'EPOLL::FD) ; EPOLL::FD) ; --> PROGN SETF LET* MULTIPLE-VALUE-BIND LET PROGN CFFI::FOREIGN-SLOT-SET ; --> SETF LET* MULTIPLE-VALUE-BIND LET PROGN CFFI::MEM-SET ; --> CFFI-SYS:%MEM-SET LET SETF ; ==> ; (SB-KERNEL:%SET-SAP-REF-SAP EPOLL::DATA-PTR 0 #:VALUE174) ; ; caught WARNING: ; Asserted type SB-SYS:SYSTEM-AREA-POINTER conflicts with derived type ; (VALUES (INTEGER 0 0) &OPTIONAL). ; See also: ; The SBCL Manual, Node "Handling of Types" ; ; compilation unit finished ; caught 1 WARNING condition ; printed 2 notes

2 1

[cffi-devel] groveller doesn't grovel for functions?
by Andy Chambers 10 Sep '09

10 Sep '09

Hi, Am I correct in thinking that the groveller doesn't grovel for functions and automatically create the correct defcfun forms for a given library? -- ---- Andy Chambers Formedix Ltd

4 5

[cffi-devel] Accept defcfun name as single element list
by Johan Ur Riise 09 Sep '09

09 Sep '09

Of these 10 forms of defcfun, the last two are illegal. They have either the name as a string or a symbol as the only element of a list. According to the documentation, they probably should be legal too. I suggest a change to allow it. (defcfun "gethostname" :int (name :pointer) (namelen :int)) (defcfun gethostname :int (name :pointer) (namelen :int)) (defcfun ("gethostname" lispgethostname) :int (name :pointer) (namelen :int)) (defcfun (lispgethostname "gethostname") :int (name :pointer) (namelen :int)) (defcfun ("gethostname" :convention :stdcall) :int (name :pointer) (namelen :int)) (defcfun (gethostname :convention :stdcall) :int (name :pointer) (namelen :int)) (defcfun (lispgethostname "gethostname" :convention :stdcall) :int (name :pointer) (namelen :int)) (defcfun ("gethostname" lispgethostname :convention :stdcall) :int (name :pointer) (namelen :int)) ;; illegal: (defcfun ("gethostname") :int (name :pointer) (namelen :int)) ;; illegal: (defcfun (gethostname) :int (name :pointer) (namelen :int)) This is from function.lisp. ;;; The following four functions take care of parsing DEFCFUN's first ;;; argument whose syntax can be one of: ;;; ;;; 1. string ;;; 2. symbol ;;; 3. \( string [symbol] options* ) ;;; 4. \( symbol [string] options* ) ;;; ;;; The string argument denotes the foreign function's name. The ;;; symbol argument is used to name the Lisp function. If one isn't ;;; present, its name is derived from the other. See the user ;;; documentation for an explanation of the derivation rules. And from the documentation of defcfun. This is a bit more clear, except that keyword parameters normally are optional, and in this case they indeed have default values >From the manual: name-and-options ::= name | (name &key library calling-convention cconv) name ::= lisp-name [foreign-name] | foreign-name [lisp-name] This is the lisp-name function Even if the system here is conformant with the documentation, it would not hurt to allow the last two forms. I suggest to change the first test from (keywordp (second spec)) to (oddp (length spec)) (defun lisp-name (spec &optional varp) (etypecase spec (list (if (keywordp (second spec)) (lisp-name (first spec) varp) (if (symbolp (first spec)) (first spec) (lisp-name (second spec) varp)))) (string (intern (format nil (if varp "*~A*" "~A") (canonicalize-symbol-name-case (substitute #\- #\_ spec))))) (symbol spec))) The foreign-option function seems to be ok as it is, as it can handle a list of length one: (defun foreign-options (spec varp) (let ((opts (if (listp spec) (if (keywordp (second spec)) (cdr spec) (cddr spec)) nil))) (if varp (funcall 'parse-defcvar-options opts) (parse-function-options opts)))) What do you think? Maybe this patch is useful - I am not yet a master of mercurial: Tue Sep 8 22:13:56 W. Europe Daylight Time 2009 Johan Ur Riise <johan(a)riise-data.no> * Accept defcfun name as single element list New patches: [Accept defcfun name as single element list Johan Ur Riise <johan(a)riise-data.no>**20090908201356 Ignore-this: 8b4f6bf7d6c7a7f6562289e931861e1d ] hunk ./src/functions.lisp 223 (defun lisp-name (spec &optional varp) (etypecase spec - (list (if (keywordp (second spec)) + (list (if (oddp (length spec)) (lisp-name (first spec) varp) (if (symbolp (first spec)) (first spec) Context: [Deprecate groveler clause FLAG in favour of CC-FLAGS. Stelian Ionescu <sionescu(a)cddr.org>**20090823121108 Ignore-this: e1537717a9e7356b208d5cd14b34ba50 ] [cffi-allegro: define long long types on 64-bit platforms Luis Oliveira <loliveira(a)common-lisp.net>**20090821210052 Patch courtesy of John Fremlin. ] [cffi-tests: fix pointer-to-integer casts in libtest.c Luis Oliveira <loliveira(a)common-lisp.net>**20090821205447] [cffi-tests: don't use stdcall #ifndef WIN32 Luis Oliveira <loliveira(a)common-lisp.net>**20090821205144] [Also canonicalize search paths in library specs. Stelian Ionescu <sionescu(a)cddr.org>**20090809005356 Ignore-this: 86a039c7ebbc757c9934fe99368ae0bb ] [Update manual. Stelian Ionescu <sionescu(a)cddr.org>**20090808222834 Ignore-this: 15e832e5220a6ca70722730d81edf283 DEFCFUN & co. now take only :CONVENTION. ] [Whitespace. Stelian Ionescu <sionescu(a)cddr.org>**20090808221547 Ignore-this: 18d99969b97b190176e88d9eb24a94ce

2 3

[cffi-devel] CFFI:FOREIGN-ALLOC Problem
by Volkan YAZICI 08 Sep '09

08 Sep '09

Hi, I use below macro to allocate foreign vectors. (defmacro with-foreign-vectors ((&rest bindings) &body body) "Convenient macro to create foreign vector bindings. `BINDIGS' are of `(NAME TYPE INITIAL-CONTENTS)' format." `(let* ,(loop with val = (gensym) for (name type initial-contents) in bindings collect `(,name (let ((,val ,initial-contents)) ; (let ((*print-length* 10)) ; (format *trace-output* "~S => ~D ~S (~A)~%" ; ',name (when ,val (length ,val)) ; ,val (find nil ,val))) (foreign-alloc ,type :count (length ,val) :initial-contents ,val)))) (unwind-protect (progn ,@body) ,@(mapcar (lambda (name) `(foreign-free ,name)) (mapcar #'car bindings))))) But while running the program, it throws below error message. (I experience same situation in both SBCL and CCL.) NALP> (compare-partitionings "/home/vy/bilkent/projects/xml/t/xmark-1024-20091006-1314-small/" "/home/vy/bilkent/projects/xml/t/xmark-1024-20091006-1314-small/sanitized-logs.out" "/home/vy/bilkent/projects/xml/t/xmark-1024-20091006-1314-small/weights.out" 40 16 0.2) N-PTR => 1 #(7536) (NIL) XADJ-PTR => 7537 #(0 1 3 5 6 8 10 12 14 16 ...) (NIL) ADJNCY-PTR => 8011 #(1 80 2 5666 3 4 2044 5 2292 6 ...) (NIL) ; Evaluation aborted. ; The value NIL is not of type (SIGNED-BYTE 32). ; [Condition of type TYPE-ERROR] ; ; Restarts: ; 0: [RETRY] Retry SLIME REPL evaluation request. ; 1: [ABORT] Return to SLIME's top level. ; 2: [TERMINATE-THREAD] Terminate this thread (#<THREAD "repl-thread" RUNNING {AA75079}>) ; ; Backtrace: ; 0: (CFFI-SYS:%MEM-SET NIL #.(SB-SYS:INT-SAP #X080C8850) :INT 16620) ; Locals: ; CFFI-SYS::OFFSET = 16620 ; CFFI-SYS::PTR = #.(SB-SYS:INT-SAP #X080C8850) ; TYPE = :INT ; CFFI-SYS::VALUE = NIL ; 1: (FOREIGN-ALLOC :INT)[:EXTERNAL] ; Locals: ; SB-DEBUG::ARG-0 = 5 ; SB-DEBUG::ARG-1 = :INT ; 2: (METIS-BIPARTITION-GRAPH-FFI #(0 1 3 5 6 8 ...) #(1 80 2 5666 3 4 ...) #(1 1 1 1 1 1 ...) #(20 7 13 1 19 20 ...) 0.2) ; Locals: ; SB-DEBUG::ARG-0 = #(0 1 3 5 6 8 ...) ; SB-DEBUG::ARG-1 = #(1 80 2 5666 3 4 ...) ; SB-DEBUG::ARG-2 = #(1 1 1 1 1 1 ...) ; SB-DEBUG::ARG-3 = #(20 7 13 1 19 20 ...) ; SB-DEBUG::ARG-4 = 0.2 When I look at the printed results, there doesn't appear any `NIL' values in any of the `:INITIAL-CONTENTS' vectors. But error tells that one of the `VALUE' arguments passed to `CFFI-SYS:%MEM-SET' is `NIL'. Below are some of the detailed macroexpansions related to `CFFI-SYS:%MEM-SET'. (define-mem-accessors (:char sb-sys:signed-sap-ref-8) (:unsigned-char sb-sys:sap-ref-8) (:short sb-sys:signed-sap-ref-16) (:unsigned-short sb-sys:sap-ref-16) (:int sb-sys:signed-sap-ref-32) (:unsigned-int sb-sys:sap-ref-32) (:long sb-sys:signed-sap-ref-word) (:unsigned-long sb-sys:sap-ref-word) (:long-long sb-sys:signed-sap-ref-64) (:unsigned-long-long sb-sys:sap-ref-64) (:float sb-sys:sap-ref-single) (:double sb-sys:sap-ref-double) (:pointer sb-sys:sap-ref-sap)) Which turns into below form if I macroexpand the `DEFINE-MEM-ACCESSORS'. ... (DEFUN %MEM-SET (VALUE PTR TYPE &OPTIONAL (OFFSET 0)) (ECASE TYPE (:CHAR (SETF (SB-SYS:SIGNED-SAP-REF-8 PTR OFFSET) VALUE)) (:UNSIGNED-CHAR (SETF (SB-SYS:SAP-REF-8 PTR OFFSET) VALUE)) (:SHORT (SETF (SB-SYS:SIGNED-SAP-REF-16 PTR OFFSET) VALUE)) (:UNSIGNED-SHORT (SETF (SB-SYS:SAP-REF-16 PTR OFFSET) VALUE)) (:INT (SETF (SB-SYS:SIGNED-SAP-REF-32 PTR OFFSET) VALUE)) (:UNSIGNED-INT (SETF (SB-SYS:SAP-REF-32 PTR OFFSET) VALUE)) (:LONG (SETF (SB-SYS:SIGNED-SAP-REF-WORD PTR OFFSET) VALUE)) (:UNSIGNED-LONG (SETF (SB-SYS:SAP-REF-WORD PTR OFFSET) VALUE)) (:LONG-LONG (SETF (SB-SYS:SIGNED-SAP-REF-64 PTR OFFSET) VALUE)) (:UNSIGNED-LONG-LONG (SETF (SB-SYS:SAP-REF-64 PTR OFFSET) VALUE)) (:FLOAT (SETF (SB-SYS:SAP-REF-SINGLE PTR OFFSET) VALUE)) (:DOUBLE (SETF (SB-SYS:SAP-REF-DOUBLE PTR OFFSET) VALUE)) (:POINTER (SETF (SB-SYS:SAP-REF-SAP PTR OFFSET) VALUE)))) ... To be honest, I need some urgent assistance. Any helps will be really appreciated. Regards.

1 2

[cffi-devel] Linking to fortran
by Mirko Vukovic 02 Sep '09

02 Sep '09

Hello, I am trying to link to a fortran 77 library. So far, I can call a subroutine without parameters, but fail otherwise. Here's where I'm at: The fortran routine is very simple: accept arguments and dump them to disk. The following version accepts argument a, and writes a,b & i to disk. The initial version had no arguments, and a was set internaly. This initial version worked. subroutine test_link (a)!a, b, i) real a,b integer i b=2.0 i=1 open (10, file='call_args.txt',status='replace') write (10,*) a, b, i close (10) return end My cffi definitions are (the routine is in the amos library): (define-foreign-library amos (:unix "libamos.so")) (use-foreign-library amos) (defctype test-link :int) ;; I got the trailing underscore by using `nm' on the library. (defcfun "test_link_" test-link (a :float)) ; (b :float) (i :int)) (test-link- 2.0)) ; 2.0 2.1 1) I get partial success. A call to test-link- returns 0 (not sure what that means), but the file contents are incorrect: -3.5797695E+21 2.000000 1 i.e, instead of 2.0 I get -3.579e21. I will be needing to call single and double reals and integers. I would also like eventually to pass single and double complex numbers. I hear from people that linking to fortran is doable, but I have not seen simple examples. Right now I am using gfortran. Eventually I will try intel fortran. Thanks for your help Mirko

3 4

[cffi-devel] is cffi darcs repo broken?
by Sam Steingold 01 Sep '09

01 Sep '09

$ darcs get http://common-lisp.net/project/cffi/darcs/cffi/ cffi Invalid repository: http://common-lisp.net/project/cffi/darcs/cffi darcs failed: failed to fetch: http://common-lisp.net/project/cffi/darcs/cffi/_darcs/inventory ExitFailure 1 darcs 1.0.9 (release) -- Sam Steingold <http://sds.podval.org>

3 2

[cffi-devel] utf-16 encoded strings in ClozureCL on a Win32 Host
by John Miller 01 Sep '09

01 Sep '09

A little background: I am trying to get RDNZL working on Clozure CL 1.4 running on my Win32 box using CFFI. It gets an "exception occurred while executing foreign code" that causes the listener to drop into the kernel debugger. All I can tell is that the error occurs in %invoke-static-member and I believe the error is because the function expects a utf-16 (wchar) string, but for some reason the encoding is not working as expected in CFFI. I went and did a little experimentation and found the following: CFFI> *default-foreign-encoding* :UTF-8 CFFI> (foreign-funcall "_wcsdup" (:string :encoding :utf-16) "foo" :string) "f" ;; That looks like the manual CFFI> (let ((*default-foreign-encoding* :utf-16)) (foreign-funcall "_wcsdup" (:string :encoding :utf-16) "foo" :string)) "" ;; Oops, foo in Sumerian cuneiform, maybe? Note that strdup won't work with utf-16 on windows. You use _wcsdup instead, apparently. CL-USER> (with-encoded-cstrs :utf-16 ((x "foo")) (#__wcsdup x)) #<A Foreign Pointer #x117C730> CL-USER> (ccl::%get-native-utf-16-cstring *) "foo" ;; Almost got it! CL-USER> (ccl::with-native-utf-16-cstr (x "foo") (#__wcsdup x)) #<A Foreign Pointer #x1171590> CL-USER> (ccl::%get-native-utf-16-cstring *) "foo" ;; Bulls-eye! Any idea on why this is happening? Many thanks, John Miller

2 1

[cffi-devel] New patches: 23-Aug-2009
by Luis Oliveira 24 Aug '09

24 Aug '09

Sun Aug 23 08:11:08 EDT 2009 Stelian Ionescu <sionescu(a)cddr.org> * Deprecate groveler clause FLAG in favour of CC-FLAGS. M ./doc/cffi-manual.texinfo -2 +2 M ./grovel/grovel.lisp +5 An updated tarball of CFFI's source can be downloaded here: http://common-lisp.net/project/cffi/tarballs/cffi-090823.tar.gz Darcsweb URL: http://common-lisp.net/cgi-bin/darcsweb/darcsweb.cgi?r=cffi-cffi;a=summary

1 0

[cffi-devel] Grovel default search path for include files
by Liam Healy 22 Aug '09

22 Aug '09

System-installed header files sometimes are in different places depending on the OS and method of installation. In particular, the file ffi.h that FSBV (http://repo.or.cz/w/fsbv.git) uses from libffi seems to be in different places on OS X depending on how it was installed (Apple native, Mac ports, etc.). Not being a user of OS X, it is difficult for me to know all the possible locations where this file might end up. Is it possible to specify that the groveler should "search system-standard locations" for the header files specified in include forms, analogous to load-foreign-library knowing where to look for system libraries, and have it find the file based on its knowledge of the places it might be for the current OS? Thank you. Liam

2 2

[cffi-devel] uffi.asd vs uffi-compat.asd
by Jeff Cunningham 22 Aug '09

22 Aug '09

I just tried to install elephant and ran into an annoying problem which could only be resolved by renaming the cffi/uffi-compat/uffi.asd to uffi-compat.asd. The problem seems to be one of identity theft, since elephant uses uffi, and it was picking up the cffi compatibility code (which does not appear to be sufficiently compatible for elephant), instead of what it expected. I also know that there has been prior discussion about this - I didn't read all of it, but found some on this list by Googling - I wasn't around the list at the time. Is there a way that this can be resolved? It seems that both libraries should be able to coexist on a system, as there are applications which expect them to be there. It also seems that UFFI has a prior claim to its own name for a system file. As it stands, everytime I do a clbuild upgrade I am going to have to make sure I fix the link. Regards, --Jeff

7 15