General discussion of Steel Bank Common Lisp

Nikodemus Siivola | 1 Mar 2007 10:47

Re: float.pure.lisp / NAN-COMPARISONS failure on Linux/x86

Stefan Lang wrote:
>  Failure:             float.pure.lisp / NAN-COMPARISONS

> If this is expected after Nikodemus Siivola's last checkin,
> ignore this mail 

No, this is not expected. Thanks for the report!

What platform does this happen on, and can you send the output from

  sh run-tests.sh float.pure.lisp

?

Cheers,

   -- Nikodemus

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

headers

Stefan Lang | 1 Mar 2007 14:01

Re: float.pure.lisp / NAN-COMPARISONS failure on Linux/x86

On Thursday 01 March 2007 10:47, Nikodemus Siivola wrote:
> Stefan Lang wrote:
> >  Failure:             float.pure.lisp / NAN-COMPARISONS
> >
> > If this is expected after Nikodemus Siivola's last checkin,
> > ignore this mail 
>
> No, this is not expected. Thanks for the report!
>
> What platform does this happen on, and can you send the output from
>
>   sh run-tests.sh float.pure.lisp

$ uname -a
Linux linux-sl 2.6.13-15-default #1 Tue Sep 13 14:56:15 UTC 2005 i686 
i686 i386 GNU/Linux

$ sh run-tests.sh float.pure.lisp &> float.pure.out

Attachment (float.pure.out): text/x-objcsrc, 7936 bytes

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

_______________________________________________
Sbcl-help mailing list
Sbcl-help <at> lists.sourceforge.net

(Continue reading)

headers

Nikodemus Siivola | 2 Mar 2007 18:05

Re: float.pure.lisp / NAN-COMPARISONS failure on Linux/x86

Stefan Lang wrote:

>>>  Failure:             float.pure.lisp / NAN-COMPARISONS

Failure was due a bad test, fixed in 1.0.3.13. Thanks for the report!

Cheers,

   -- Nikodemus

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

headers

Mark W Johnson | 3 Mar 2007 08:27

sb-alien array allocation

I would like to be able to generate alien arrays whose length is determined at run time, but I am having trouble understanding how to do this using (sb-alien:with-alien () ) or make-alien.

My attempt at this, along the lines shown below, fails:

> This is SBCL 1.0.2, an implementation of ANSI Common Lisp.
> More information about SBCL is available at < http://www.sbcl.org/>.
> SBCL is free software, provided as is, with absolutely no warranty.
...

> CL-USER> (defun my-func (a-seq)
>          (sb-alien:with-alien
>           ((x (array (sb-alien:unsigned 32) (length a-seq)))) x))
>
>; in: LAMBDA NIL
>;     (WITH-ALIEN ((X (ARRAY (UNSIGNED 32) (LENGTH A-SEQ)))) X)
>;
>; caught ERROR:
>;   (in macroexpansion of (WITH-ALIEN # X))
>;   (hint: For more precise location, try *BREAK-ON-SIGNALS*.)
>;   The first dimension is not a non-negative fixnum or NIL: (LENGTH A-SEQ)

>;     (SB-INT:NAMED-LAMBDA MY-FUNC (A-SEQ) (BLOCK MY-FUNC (WITH-ALIEN ((X #)) X)))
>; ==>
>;   #'(SB-INT:NAMED-LAMBDA MY-FUNC (A-SEQ) (BLOCK MY-FUNC (WITH-ALIEN ((X #)) X)))
>;
>; caught STYLE-WARNING:
>;   The variable A-SEQ is defined but never used.
>;
>; compilation unit finished
>;   caught 1 ERROR condition
>;   caught 1 STYLE-WARNING condition
...

I find I can only use a literal fixnum, not a symbol or form that evaluates to a fixnum, in the dimension specification for the array x.

A second attempt (which also includes a simple test):

> (defun my-func (a-seq)
>       (eval
>        `(sb-alien:with-alien
>          ((x (array (sb-alien:unsigned 32) ,(length a-seq))))
>          (loop for k below ,(length a-seq) do
>            (setf (sb-alien:deref x k) (elt ,a-seq k))
>            (format t "~d:~d~%" k (sb-alien:deref x k))))))
>MY-FUNC
>CL-USER> (my-func #(1 2 3 4 5 4 3 2 1))
>0:1
>1:2
>2:3
>3:4
>4:5
>5:4
>6:3
>7:2
>8:1
>NIL
>
appears to work, but seems awkward to me - as though I am missing something. Is there a more straightforward way to go about this?
The point of converting to an alien array is for passing data to a linux kernel module. At the moment I am using a solution (not shown) that simply hard-wires the array size to the maximum possible for the application, and this works, but I am hoping for something more efficient.

In case it matters, I am using sbcl 1.02 on
Linux 2.6.17-1.2174_FC5 #1 Tue Aug 8 15:30:55 EDT 2006 i686 i686 i386 GNU/Linux

Regards,
Mark

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

_______________________________________________
Sbcl-help mailing list
Sbcl-help <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-help

headers

Juho Snellman | 3 Mar 2007 10:06

Re: sb-alien array allocation

"Mark W Johnson" <mwjohnson0 <at> gmail.com> writes:
> I would like to be able to generate alien arrays whose length is determined
> at run time, but I am having trouble understanding how to do this using
> (sb-alien:with-alien () ) or make-alien.

For MAKE-ALIEN you can specify the size dynamically with the optional
SIZE parameter. Just use a dummy value for the dimension in the
type. For example (make-alien (array (unsigned 32) 0) y).

--

-- 
Juho Snellman

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

headers

Mark W Johnson | 3 Mar 2007 22:34

Re: sb-alien array allocation

This is just what I needed.
Thank You.

Mark

On 03 Mar 2007 11:06:47 +0200, Juho Snellman <jsnell <at> iki.fi > wrote:

"Mark W Johnson" <mwjohnson0 <at> gmail.com > writes:
> I would like to be able to generate alien arrays whose length is determined
> at run time, but I am having trouble understanding how to do this using
> (sb-alien:with-alien () ) or make-alien.

For MAKE-ALIEN you can specify the size dynamically with the optional
SIZE parameter. Just use a dummy value for the dimension in the
type. For example (make-alien (array (unsigned 32) 0) y).

--
Juho Snellman

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

_______________________________________________
Sbcl-help mailing list
Sbcl-help <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-help

headers

Edi Weitz | 5 Mar 2007 13:04

Useless backtraces / functions called NIL

I'm currently working on parts of a large project that I don't have
complete control of.  I don't even know all parts of it good enough
yet.  What I /do/ know, though, is that I sometimes (pretty often,
actually) get completely useless backtraces where the "interesting"
function calls are replaced with calls to functions named NIL.  I've
attached a part of such a backtrace below (some names modified because
the code is confidential).  The frames I'm interested in in this case
are #13 and #14 - note that they both start with NIL as well as the
one in frame #16.

When the functions appearing in frames 13 to 16 where compiled (with
ASDF), SB-C::*POLICY* looked like this

  ((SB-EXT:INHIBIT-WARNINGS . 1)
   (SPEED . 1)
   (SAFETY . 3)
   (COMPILATION-SPEED . 1)
   (SPACE . 1)
   (DEBUG . 3))

which I think is not a very aggressive setting.  So, what could be the
reason for this behaviour?  Any other settings that might affect this?
I skimmed the list archives a bit, but I didn't really know what to
search for.

I suppose/hope it is not expected that I see these "NIL frames" and
that I can make them go away somehow.

Thanks in advance for your help,
Edi.

  7: (SB-KERNEL::ODD-KEY-ARGS-ERROR-HANDLER
      #<unavailable argument>
      #<unavailable argument>
      #<unavailable argument>
      #<unavailable argument>)
  8: (SB-KERNEL:INTERNAL-ERROR
      #.(SB-SYS:INT-SAP #XF63C3EDC)
      #<unavailable argument>)
  9: ("foreign function: call_into_lisp")
  10: ("foreign function: funcall2")
  11: ("foreign function: interrupt_internal_error")
  12: ("foreign function: sigtrap_handler")
  13: (NIL :PARTS)
  14: (NIL
       #<FOO-OBJECT>
       #<BAR-OBJECT>
       #<FROB-OBJECT>
       :SOME-KEYWORD)
  15: (DO-SOMETHING
       #<QUUX-OBJECT>
       NIL)
  16: (NIL FROB-TEST)
  17: (SB-INT:SIMPLE-EVAL-IN-LEXENV
       (RUN-TEST 'FROB-TEST)
       #<NULL-LEXENV>)
  18: (SWANK::EVAL-REGION
       "(run-test 'frob-test)
  "
       T)

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

headers

Juho Snellman | 5 Mar 2007 14:59

Re: Useless backtraces / functions called NIL

Edi Weitz <edi <at> agharta.de> writes:
> I'm currently working on parts of a large project that I don't have
> complete control of.  I don't even know all parts of it good enough
> yet.  What I /do/ know, though, is that I sometimes (pretty often,
> actually) get completely useless backtraces where the "interesting"
> function calls are replaced with calls to functions named NIL.

SBCL almost always stores enough debug information to show at least
the name of the function, regardless of the compiler policy. The only
circumstance I know of where NIL shows up as the function name in a
backtrace is for anonymous functions compiled using (compile nil ...),
or some other construct that essentially boils down to that. For
example (eval '(foo)) -> (funcall (compile nil '(lambda () (foo)))).

But it shouldn't really be happening for named functions, or when
using file compilation.

> I suppose/hope it is not expected that I see these "NIL frames" and
> that I can make them go away somehow.

Hard to say without seeing an example. The attached patch fixes the
issue I described above, but might not help if it's something else.

Index: src/compiler/main.lisp
===================================================================
RCS file: /cvsroot/sbcl/sbcl/src/compiler/main.lisp,v
retrieving revision 1.124
diff -u -r1.124 main.lisp
--- src/compiler/main.lisp	4 Mar 2007 23:47:58 -0000	1.124
+++ src/compiler/main.lisp	5 Mar 2007 13:53:12 -0000
 <at>  <at>  -976,7 +976,8  <at>  <at> 
     (let* ((locall-fun (let ((*allow-instrumenting* t))
                          (funcall #'ir1-convert-lambdalike
                                   definition
-                                  :source-name name)))
+                                  :source-name name
+                                  :debug-name definition)))
            (debug-name (debug-name 'tl-xep name))
            ;; Convert the XEP using the policy of the real
            ;; function. Otherwise the wrong policy will be used for


--

-- 
Juho Snellman

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

_______________________________________________
Sbcl-help mailing list
Sbcl-help <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sbcl-help

headers

Christophe Rhodes | 5 Mar 2007 15:33

Re: float.pure.lisp / NAN-COMPARISONS failure on Linux/x86

Nikodemus Siivola <nikodemus <at> random-state.net> writes:

> Stefan Lang wrote:
>
>>>>  Failure:             float.pure.lisp / NAN-COMPARISONS
>
> Failure was due a bad test, fixed in 1.0.3.13. Thanks for the report!

I'm still getting failures on x86-64/linux, 1.0.3.30:

  (sb-int:with-float-traps-masked (:invalid) 
    (funcall (lambda () (>= 1.0 (/ 0.0 0.0)))))

returns T, whereas 

  (sb-int:with-float-traps-masked (:invalid) 
    (eval '(>= 1.0 (/ 0.0 0.0))))

returns NIL.

Furthermore, I think there are more problems lurking; for instance,

  (lambda (x) (= x x))

will bogusly return T on a NaN argument, given the transform's use of
REFLEXIVE-P (src/compiler/srctran.lisp).  (Similarly, (/= x x)
probably should return T for x being NaN...)

Cheers,

Christophe

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

headers

Edi Weitz | 5 Mar 2007 15:48

Re: Useless backtraces / functions called NIL

On 05 Mar 2007 15:59:09 +0200, Juho Snellman <jsnell <at> iki.fi> wrote:

> SBCL almost always stores enough debug information to show at least
> the name of the function, regardless of the compiler policy. The
> only circumstance I know of where NIL shows up as the function name
> in a backtrace is for anonymous functions compiled using (compile
> nil ...), or some other construct that essentially boils down to
> that. For example (eval '(foo)) -> (funcall (compile nil '(lambda ()
> (foo)))).
>
> But it shouldn't really be happening for named functions, or when
> using file compilation.

In my case these were named functions defined with DEFUN and compiled
with COMPILE-FILE (from ASDF)... :(

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

10	April 2013
17	March 2013
8	February 2013
9	January 2013
4	December 2012
14	November 2012
10	October 2012
10	September 2012
12	August 2012
40	July 2012
13	June 2012
58	May 2012
16	April 2012
17	March 2012
22	February 2012
9	January 2012
21	December 2011
21	November 2011
25	October 2011
21	September 2011
38	August 2011
15	July 2011
28	June 2011
17	May 2011
31	April 2011
5	March 2011
31	February 2011
17	January 2011
19	December 2010
41	November 2010
33	October 2010
43	September 2010
31	August 2010
43	July 2010
24	June 2010
8	May 2010
59	April 2010
55	March 2010
44	February 2010
26	January 2010
31	December 2009
46	November 2009
59	October 2009
52	September 2009
35	August 2009
48	July 2009
58	June 2009
29	May 2009
15	April 2009
18	March 2009
31	February 2009
33	January 2009
83	December 2008
42	November 2008
52	October 2008
19	September 2008
52	August 2008
24	July 2008
25	June 2008
1	May 2008
13	April 2008
28	March 2008
17	February 2008
36	January 2008
59	December 2007
9	November 2007
16	October 2007
11	September 2007
24	August 2007
31	July 2007
31	June 2007
60	May 2007
40	April 2007
50	March 2007
37	February 2007
41	January 2007
69	December 2006
20	November 2006
39	October 2006
51	September 2006
23	August 2006
11	July 2006
29	June 2006
19	May 2006
26	April 2006
19	March 2006
37	February 2006
45	January 2006
24	December 2005
30	November 2005
17	October 2005
28	September 2005
26	August 2005
33	July 2005
37	June 2005
69	May 2005
65	April 2005
39	March 2005
32	February 2005
24	January 2005
27	December 2004
80	November 2004
20	October 2004
38	September 2004
14	August 2004
23	July 2004
15	June 2004
18	May 2004
69	April 2004
21	March 2004
15	February 2004
32	January 2004
19	December 2003
14	November 2003
3	October 2003
27	September 2003
28	August 2003
3	July 2003
10	June 2003
27	May 2003
14	March 2003
4	February 2003
5	January 2003
8	December 2002
3	September 2002
2	August 2002
5	May 2002

Mon	Tue	Wed	Thu	Fri	Sat	Sun

			1	2	3	4
5	6	7	8	9	10	11
12	13	14	15	16	17	18
19	20	21	22	23	24	25
26	27	28	29	30	31