gmane.lisp.scheme.stklos

STklos sometimes fails it's test suite

Vitaly Magerya — 2012-05-07T17:06:30

Hi Erick, hi everyone. It appears that STklos is subtly broken, as it
fails it's own test suite, but it only does so once in a while.

Here's how I build it (this is so you could easily reproduce the problem
by just copy-pasting this message to the command line):

    wget "http://www.stklos.net/download/stklos-1.10.tar.gz"
    tar xf stklos-1.10.tar.gz
    cd stklos-1.10
    ./configure --enable-threads=posix --with-provided-gc
    make
    make test

This may work at first, but if you'll repeat "make test" a few times
(20 times is usually enough), you'll see that some tests may fail. In
particular, on different occasions I've seen 3 errors:


==== Testing call/cc & dynamic-wind  ...                         failed
Errors found in this section:
test R5RS example on (let ((path '()) (c #f)) (let ((add (lambda (s)
(set! path (cons s path))))) (dynamic-wind (lambda () (add 'connect))
(lambda () (add (call-with-current-continuation (lambda (c0) (set! c
c0) 'talk1)))) (lambda () (add 'disconnect))) (if (< (length path) 4)
(c 'talk2) (reverse path)))) expected (connect talk1 disconnect
connect talk2 disconnect) but got #(fail)

==== Testing call/cc & dynamic-wind  ...                         failed
Errors found in this section:
test coroutines on (with-output-to-string start) expected
"abcdabcdabcdabcdabcdabcdabcdabcdabcdabcd" but got #(fail)

==== Testing call/cc & dynamic-wind  ...
make[1]: *** [test] Segmentation fault (core dumped)


This is what I see on Ubuntu 12.04 i386. Strangely enough, I don't see
this on a FreeBSD 9.0 amd64 system.

Next, I tried updating the bundled libgc from 7.2alpha2 (which STklos
currently ships with) to 7.2alpha{4,6} and to the latest git sources
found at [1,2]. This is how I do it (again, for your convenience):

    wget "http://www.stklos.net/download/stklos-1.10.tar.gz"
    wget "http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gc-7.2alpha6.tar.gz"
    tar xf stklos-1.10.tar.gz
    tar xf gc-7.2alpha6.tar.gz
    rm -rf stklos-1.10/gc
    mv gc-7.2alpha6 stklos-1.10/gc
    cd stklos-1.10
    ./configure --enable-threads=posix --with-provided-gc
    make
    make test

(You can use [3] to get the snapshot of the current libgc + libatomic_ops
sources which I used).

So, on Ubuntu I see no difference: STklos with every libgc from 7.1 to
latest git exhibit the same problem. On FreeBSD I see no such problems
with libgc 7.1 and 7.2alpha2, but 7.2alpha{4,6} and current git are
all affected in a similar way (there's additionally a hang on threading
tests).

(Unfortunately starting with FreeBSD 9.0 libgc 7.1 and 7.2alpha2 crash
on startup if you link with them dynamically; because of this STklos
compiled from ports collection does not currently work on FreeBSD 9.0;
I intend to fix this as soon as this problem is resolved).

So, can you reproduce the problem? Can you fix it?

[1] https://github.com/ivmai/bdwgc
[2] https://github.com/ivmai/libatomic_ops
[3] http://tx97.net/pub/distfiles/boehm-gc-7.3.a1.20120502.tar.xz

Re: Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-02-08T02:57:52

(Sorry I only just now got back around to this email.)

OK, I see. Thank you very much for clarifying that.

Re: STklos failes to run Silex without reporting an error

Erick Gallesio — 2012-01-18T22:19:13

Sascha,

I have never seen this problem before (and Silex is used in the build of 
STklos). BTW , Silex and the lalr parser of Dominique Boucher are 
present in STklos distribution since a long time. But I must admit that 
I have probably never tried to build an automaton of this size.

STklos provides a command to build an automaton which is stklos-genlex. 
It produces an automaton which is not a table, but a program. It crashes 
also with your ASN1 grammar (the automaton is created but the lexer 
closure cannot be built).

The number of arguments should not a problem, as exercised below:

    stklos> (define (iota n)
      (letrec ((aux (lambda (n l) (if (>= n 0) (aux (- n 1) (cons n l)) 
l))))
        (aux (- n 1) '())))
    ;; iota
    stklos> (length (apply append (map list (iota 20000))))
    20000

I'll try to investigate the problem, but it could not be before some 
days. In the meantime a simple way to make the automaton smaller could 
be to not build a rule for all the reserved identifiers and recognize 
them in the identifier rule. Something like

     {identifier} (let ((reserved (member yytext '("ABSENT" "ALL" ....))))
                            (if reserved
                                (token (string->symbol (car reserved)))
                                (token 'identifier yytext)))

shouldl reduce drastically the size of the automaton.

STklos failes to run Silex without reporting an error

Sascha Ziemann — 2012-01-17T15:59:28

I am trying to use Silex with STklos. Silex is a lexer generator for Scheme:

http://www.iro.umontreal.ca/~dube/silex.tar.gz

But when I run Silex with STklos it fails to produce the lexer table
without reporting any error:

$ stklos
*   STklos version 1.10
 *  Copyright (C) 1999-2011 Erick Gallesio - Universite de Nice <eg< at >unice.fr>
* * [Linux-2.6.32-bpo.5-amd64-x86_64/pthread/readline/utf8]
stklos> (load "silex/silex.scm")
stklos> (lex "asn1.l" "stklos.scm")
#t
stklos>

The above session creates the file stklos.scm which contains the
lexer. I have attached the file. You can find at line 1756 a single
quote without any value. That is the bug.

Sorry for the big mail but the bug occurs only with big lexers.

When I run Silex with Gauche it works as expected. I have attached the
lexer generated by Gauche, which contains the correct value behind the
quote.

I have also attached my grammar file in order to reproduce the bug.

Silex stresses the Scheme interpreter by the number of arguments
passed to string-append during an (apply string-append lst) with a lst
containing more than 8k elements. Gambit failed already on this:
https://mercure.iro.umontreal.ca/pipermail/gambit-list/2012-January/005560.html

Is it possible that STklos has a similar problem?
_______________________________________________
STklos mailing list
STklos< at >email.essi.fr
http://email.essi.fr/mailman/listinfo/stklos

Re: Objects & SRFI-10: No Applicable Method

Erick Gallesio — 2012-01-15T21:44:57

Taylor,

My previous response was incorrect: the current behavior is the one 
which is defined by SRFI-10 which insists on the difference of the 
Scheme SRFI-10 '#,' mechanism with the CL '#.' one. It is said:

     Common Lisp defines an external form #.obj, which instructs the 
Lisp reader to
     evaluate obj right after the reader parsed it. While #. is a 
general-purpose
     read-evaluator:
           #.obj === (eval obj)
     #, is merely an application:
           #,obj === (apply (lookup (car obj)) (cdr obj))
     and obj must be an external representation of a list. Read-time 
application
     is weaker than a generic read-time evaluation, and therefore less
      expensive. Example 5 below illustrates the difference.

where Example 5 is:

     Differences between Common Lisp's #. and the proposed #,

        (with-input-from-string "#,(+ 1 (+ 2 3))" read) ==>
            error: can't add number 1 and a list '(+ 2 3)
        (with-input-from-string "#,(+ 1 #,(+ 2 3))" read) ==> 6

     In contrast, in Common Lisp

        (with-input-from-string (is "#.(+ 1 (+ 2 3))") (read is))  ==> 6

So, in your example <foo> is a symbol and not the <foo> class (in
STklos, classes are "first class" (!!) and the fist parameter of 
make-instance
must be a class object, not a symbol).

Your solution is then correct, and I'm afraid that the eval cannot be 
avoided.

I'll correct STklos documentation to take into account the given example 
which uses the '#,' syntax.

Re: Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-01-10T02:23:04

Thanks for the reply. For what it's worth, I found a workaround;
throwing in an eval makes it work:

stklos> (define-class <foo> () ())
;; <foo>
stklos> (define-reader-ctor 'make (lambda x (eval `(make ,< at >x))))
stklos> (with-input-from-string "#,(make <foo>)" read)
#[<foo> 1bd6f30]

Re: Objects & SRFI-10: No Applicable Method

Erick Gallesio — 2012-01-09T23:46:28

Hi Taylor,

Welcome on STklos. That's weird. It seems to be a regression.
I'll have a look at it (probably not before a couple of days) and post a 
correction.

Bye

Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-01-08T22:30:18

Hi there, I'm new to STklos and am trying to write my first little
program. I want to store objects into a file and read them out easily,
so looking at the manual I found an example of using write-object to
print out a representation ala SRFI-10. When I try this, however, it
doesn't work as I expected.

*   STklos version 1.10
 *  Copyright (C) 1999-2011 Erick Gallesio - Universite de Nice <eg< at >unice.fr>
* * [Linux-3.1.6-1-ARCH-x86_64/pthread/readline/utf8]
stklos> (define-class <foo> () ())
;; <foo>
stklos> (describe (make <foo>))
#[<foo> 1c00db0] is an an instance of class <foo>.
Slots are:
stklos> (define-reader-ctor 'make make)
stklos> (with-input-from-string "#,(make <foo>)" read)
**** Error:
error: no applicable method for #[<generic> make-instance (1)]
in call (make-instance <foo>)
        (type ",help" for more information)
stklos> (make-instance <foo>)
#[<foo> 1c2c240]

What am I doing wrong? Thanks!

--
Taylor C. Venable
http://metasyntax.net/
_______________________________________________
STklos mailing list
STklos< at >email.essi.fr
http://email.essi.fr/mailman/listinfo/stklos

Re: Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-01-10T02:23:04

Thanks for the reply. For what it's worth, I found a workaround;
throwing in an eval makes it work:

stklos> (define-class <foo> () ())
;; <foo>
stklos> (define-reader-ctor 'make (lambda x (eval `(make ,< at >x))))
stklos> (with-input-from-string "#,(make <foo>)" read)
#[<foo> 1bd6f30]

Re: Objects & SRFI-10: No Applicable Method

Erick Gallesio — 2012-01-09T23:46:28

Hi Taylor,

Welcome on STklos. That's weird. It seems to be a regression.
I'll have a look at it (probably not before a couple of days) and post a 
correction.

Bye

Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-01-08T22:30:18

Hi there, I'm new to STklos and am trying to write my first little
program. I want to store objects into a file and read them out easily,
so looking at the manual I found an example of using write-object to
print out a representation ala SRFI-10. When I try this, however, it
doesn't work as I expected.

*   STklos version 1.10
 *  Copyright (C) 1999-2011 Erick Gallesio - Universite de Nice <eg< at >unice.fr>
* * [Linux-3.1.6-1-ARCH-x86_64/pthread/readline/utf8]
stklos> (define-class <foo> () ())
;; <foo>
stklos> (describe (make <foo>))
#[<foo> 1c00db0] is an an instance of class <foo>.
Slots are:
stklos> (define-reader-ctor 'make make)
stklos> (with-input-from-string "#,(make <foo>)" read)
**** Error:
error: no applicable method for #[<generic> make-instance (1)]
in call (make-instance <foo>)
        (type ",help" for more information)
stklos> (make-instance <foo>)
#[<foo> 1c2c240]

What am I doing wrong? Thanks!

--
Taylor C. Venable
http://metasyntax.net/
_______________________________________________
STklos mailing list
STklos< at >email.essi.fr
http://email.essi.fr/mailman/listinfo/stklos

Objects & SRFI-10: No Applicable Method

Taylor Venable — 2012-01-08T22:30:18

Hi there, I'm new to STklos and am trying to write my first little
program. I want to store objects into a file and read them out easily,
so looking at the manual I found an example of using write-object to
print out a representation ala SRFI-10. When I try this, however, it
doesn't work as I expected.

*   STklos version 1.10
 *  Copyright (C) 1999-2011 Erick Gallesio - Universite de Nice <eg< at >unice.fr>
* * [Linux-3.1.6-1-ARCH-x86_64/pthread/readline/utf8]
stklos> (define-class <foo> () ())
;; <foo>
stklos> (describe (make <foo>))
#[<foo> 1c00db0] is an an instance of class <foo>.
Slots are:
stklos> (define-reader-ctor 'make make)
stklos> (with-input-from-string "#,(make <foo>)" read)
**** Error:
error: no applicable method for #[<generic> make-instance (1)]
in call (make-instance <foo>)
        (type ",help" for more information)
stklos> (make-instance <foo>)
#[<foo> 1c2c240]

What am I doing wrong? Thanks!

--
Taylor C. Venable
http://metasyntax.net/
_______________________________________________
STklos mailing list
STklos< at >email.essi.fr
http://email.essi.fr/mailman/listinfo/stklos

[Announce] STklos-1.10 release

Erick Gallesio — 2011-12-31T17:14:54

This is an announce for the 1.10 release of STklos.


STklos is a free Scheme system compliant with the languages features
defined in R5RS. The aim of this implementation is to be fast as well
as light. The implementation is based on an ad-hoc Virtual
Machine. STklos can also be compiled as a library and embedded in an
application.

The salient points of STklos are:

    * an efficient and powerful object system based on CLOS 
      providing
          o Multiple Inheritance,
          o Generic Functions,
          o Multi-methods
          o an efficient MOP (Meta Object Protocol)
    * a simple to use module system
    * a full tower of numbers implementation, as defined in R5RS,
    * easy connection to the GTK+ toolkit,
    * a Perl compatible regular expressions thanks to the PCRE 
      package,
    * easily extensible with its ScmPkg mechanism,
    * it implements properly tail recursive calls.

This version brings the support of UTF-8 encoding and begins to support some (future) R7RS traits:

    * Added full support for UTF-8 strings and symbols
    * Added support for Win32 support under Cygwin
    * Added some support for R7RS traits:
        o bytesvectors
        o hexadecimal chars in strings and symbols
        o continuations lines in strings
        o New primitives
            - char-foldcase
            - string-foldcase
            - string-foldcase!
            - make-list
            - string-map
            - string-for-each
            - vector-for-each
            - vector-map
    * Added support for MacOs X Lion
    * Added back support for Win32 using Cygwin (this support has still rough
      edges. See the file PORTING-NOTES in the distribution directory for
      details)
    * Bug Fixes

Homepage/Download: See http://www.stklos.net/

STklos update

Erick Gallesio — 2011-08-28T17:37:39

Hi,

I have committed a large set of modifications to STklos that were in a 
private development tree. The modifications permits mainly to support 
UTF-8 encoding in strings. There are also patches to support the future 
R7RS version of Scheme (small language). There are some things that 
still need to be done to have a new version, but current code should be 
usable. If you don't want to use UTF-8 encoding, you can use the 
--utf8-encoding=no option.

If you are interested, the code is available at 
http://stklos.net/hg/hgweb.cgi/archive/tip.tar.gz

A bientôt

FFI docs and examples

nitralime — 2011-01-28T11:22:39

Hi folks!

Is there a better (resp. more detailed) documentation of FFI?
Do you know examples of binding external libraries which
can be used to get familiar with FFI?

Regards
Nik
_______________________________________________
STklos mailing list
STklos< at >email.essi.fr
http://email.essi.fr/mailman/listinfo/stklos

Re: parameterize

Erick Gallesio — 2011-01-07T08:55:16

Hi Joo,

You are correct. I will try to fix that asap, but I suspect that it will 
not be so easy.
Thanks for signaling this bug.

parameterize

Joo ChurlSoo — 2011-01-07T00:42:42

The following parameterize behavior seems to be wrong.

(let ((f (make-parameter 'a))
      (path '())
      (c #f))
  (let ((add (lambda () (set! path (cons (f) path)))))
    (add)
    (parameterize ((f 'b))
   (call-with-current-continuation (lambda (c0) (set! c c0)))
   (add) (f 'c) (add))
    (f 'd)
    (add)
    (if (< (length path) 5)
 (c 'end)
 (reverse path))))
 (a b c d b c d) ;This answer seems to be wrong. (cf.) (a b c d c c d)

[Announce] STklos-1.01 release

2010-12-30T19:04:57

This is an announce for the 1.01 release of STklos.


STklos is a free Scheme system compliant with the languages features
defined in R5RS. The aim of this implementation is to be fast as well
as light. The implementation is based on an ad-hoc Virtual
Machine. STklos can also be compiled as a library and embedded in an
application.

The salient points of STklos are:

    * an efficient and powerful object system based on CLOS 
      providing
          o Multiple Inheritance,
          o Generic Functions,
          o Multi-methods
          o an efficient MOP (Meta Object Protocol)
    * a simple to use module system
    * a full tower of numbers implementation, as defined in R5RS,
    * easy connection to the GTK+ toolkit,
    * a Perl compatible regular expressions thanks to the PCRE 
      package,
    * easily extensible with its ScmPkg mechanism,
    * it implements properly tail recursive calls.

This release is mainly a maintenance version. Changes in this version:
     * Documentation update
     * Modifications to support ScmPkg on MacOs MacPorts
     * Fix incompatibilities in FFI introduced in 1.00
     * Better x86_64 support
     * Bug fixes

Homepage/Download: See http://www.stklos.net/

Re: Nested quasiquotation bug

Erick Gallesio — 2010-12-15T20:56:19

Hi,

You are right, this is a bug. I will try to fix it asap (proabably not 
before the holidays, unfortunately).

Nested quasiquotation bug

Rock — 2010-12-14T23:48:17

I think I've discovered a bug here.

The following:

``(,< at >(append ,< at >'((list 2 3) (list 5 7))))

should produce:

(quasiquote ((unquote-splicing (append (list 2 3) (list 5 7)))))

as it does correctly with Guile.

But with stklos 1.0, you get:

`(,< at >(append ,< at >'((list 2 3) (list 5 7))))

As you can see, it doesn't evaluate the internal unquote-splicing at all!

Thank you.

Internat d'Été?

Michael Feyereisen — 2010-12-11T18:20:33

Cher Erick,

STklos m'interesse beaucoup, et j'aimerais savoir s'il serait possible
de faire un internat à l'Université de Nice cet été, afin par exemple
de vous aider à écrire de la documentation, voire du code (en Scheme
ou en C).

Merci beaucoup,

Michael Feyereisen

Imperial College London
Department of Physics