[ clisp-Bugs-3086793 ] double-float-epsilon (and negative) invariants don't hold

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Bugs item #3086793, was opened at 2010-10-13 17:25
Message generated for change (Comment added) made by seh
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101355&aid=3086793&group_id=1355

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: clisp
Group: ANSI compliance issue
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Steven Harris (seh)
Assigned to: Bruno Haible (haible)
Summary: double-float-epsilon (and negative) invariants don't hold

Initial Comment:
This problem looks similar to bug 501969, closed back in 2002.

CLISP violates the contracts specified in the Hyperspec for double-float-epsilon and double-float-negative-epsilon, though it does work fine for single-float-epsilon and single-float-negative-epsilon.
http://www.lispworks.com/documentation/HyperSpec/Body/v_short_.htm

CL-USER> (not (= (float 1 double-float-epsilon)
                         (+ (float 1.0d0 double-float-epsilon)
                             double-float-epsilon)))
NIL
CL-USER> (not (= (float 1 double-float-negative-epsilon)
                         (- (float 1 double-float-negative-epsilon)
                         double-float-negative-epsilon)))
NIL

I'm using the CLISP binary supplied by the Cygwin project on Windows XP.

% uname -a
CYGWIN_NT-5.1 PA10-5ZYH3M1 1.7.7(0.230/5/3) 2010-08-31 09:58 i686 Cygwin

% clisp --version
GNU CLISP 2.48 (2009-07-28) (built on ATGRZWN502840.avl01.avlcorp.lan [157.247.26.41])
Software: GNU C 4.3.4 20090804 (release) 1
gcc-4 -O2 -pipe  -g -W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type -Wmissing-declarations -Wno-sign-compare -Wno-format-nonliteral -O2 -fexpensive-optimizations -falign-functions=4 -DUNICODE -DDYNAMIC_FFI -I. -Wl,--stack,8388608 -Wl,--enable-auto-import  /usr/lib/libintl.dll.a /usr/lib/libiconv.dll.a /usr/lib/libreadline.dll.a -lncurses /usr/lib/libavcall.a /usr/lib/libcallback.a /usr/lib/libiconv.dll.a -L/usr/lib -lsigsegv
SAFETY=0 HEAPCODES STANDARD_HEAPCODES SPVW_PAGES SPVW_MIXED
libsigsegv 2.8
libiconv 1.13
libreadline 6.0
Features:
(READLINE REGEXP SYSCALLS I18N LOOP COMPILER CLOS MOP CLISP ANSI-CL COMMON-LISP
 LISP=CL INTERPRETER SOCKETS GENERIC-STREAMS LOGICAL-PATHNAMES SCREEN FFI
 GETTEXT UNICODE BASE-CHAR=CHARACTER PC386 UNIX CYGWIN)
C Modules: (clisp i18n syscalls regexp readline)
Installation directory: /usr/lib/clisp-2.48/
User language: ENGLISH
Machine: I686 (I686) PA10-5ZYH3M1.gddsi.com [10.245.57.23]

----------------------------------------------------------------------

>Comment By: Steven Harris (seh)
Date: 2010-10-13 23:27

Message:
Thank you, Sam. I'll have to wait for a few days before I can post to the
Cygwin mailing list, but I'll inquire there next.

By the way, I had tried to post a comment relating this problem to a
similar discussion on the CMUCL mailing list from August 2003. Perhaps my
comment was not saved properly. Here are the references again:

http://osdir.com/ml/lisp.cmucl.devel/2003-08/msg00226.html
http://osdir.com/ml/lisp.cmucl.devel/2003-08/threads.html

----------------------------------------------------------------------

Comment By: Sam Steingold (sds)
Date: 2010-10-13 17:44

Message:
I cannot reproduce the problem either on linux or mingw or cygwin
(CYGWIN_NT-5.2 ... 1.5.25(0.156/4/2) 2008-06-12 19:34 i686 Cygwin) either
with cvs head or 2.48.
the forms above return T.
actually, this is tested by the regression test suite , so the cygwin
clisp maintainer should have noticed the problem before he distributed the
package.
the only difference I see between your setup and mine is that you have
cyginw 1.7 on winXP and I have cygwin 1.5 on server 2003.
I think it would be a good idea to ask the clisp cygwin maintainer about
this (ask on the cygwin mailing list).

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101355&aid=3086793&group_id=1355