[Caml-list] Problem with the old Num implementation of ocaml (fwd)

[Thomas Fischbacher <Thomas.Fischbacher@Physik.Uni-Muenchen.DE>]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 4863 bytes --]


The following pieces of email I just submitted to the Debian Bug Tracking 
System; Sven Luther (the Debian Maintainer) recommended to also forward it 
to this list, as this may be of interest to more people...

---------- Forwarded message ----------
Date: Thu, 18 Dec 2003 21:15:40 +0100 (CET)
From: Thomas Fischbacher <tf@cip.physik.uni-muenchen.de>
To: submit@bugs.debian.org
Subject: Problem with the old Num implementation of ocaml


Package:  ocaml 
Version: 3.06-21
Severity: grave


Please note that this bug concerns the "old" Num implementation that was 
removed in a recent update. I nevertheless use(d) it as I am right 
now doing research for a new string theory paper and wanted to 
consistently use one and the same somewhat environment throughout the 
process. Nevertheless, I suppose this affects all versions having the old num 
implementation (next thing I'll do is to upgrade my entire system to the 
new implementation and check the behaviour of that one for some large test 
calculations).

The attachment will reproduce a condition where doing the same calculation 
(which is purely functional to the outside and does not use global state 
or side effects, but involves Num) many times in sequence will give 
different results.

As you see, about half of this file is data; I did not manage to find a 
smaller test case creating that condition. Although I consider this highly
unlikely, this piece of code may involve an "int"-type related integer 
overflow. (Nevertheless, even so, subsequent identical calculations should 
not give different results...)


On my system(*), the output of a num_ocaml interpreter built as described 
in the Num section of the documentation, but also including the Unix 
module is:

0 ==> -12747938802280433395210374808335212773156996019062242486412495785350259890027861091589626807758181401556588623046875/764762496031801254357991532266868085634832685275244705613839817948600630474284529321916381431208558769420593384116260078281979735715302794541235545562947747053568
1 ==> -24799801902139948285742077521010897923261300584495629346339410063333039539280530883995185222932342824043982837509521484375/4156317913216311164989084414493848291493655898235025573988259880155438209099372441966936855604394341138155398826718804773271628998452732579028454051972542103552
2 ==> -39148920061803210956691061036397438426365134774540146675772774556810648122275561412271743926625375084180283661376953125/2294287488095403763073974596800604256904498055825734116841519453845801891422853587965749144293625676308261780152348780234845939207145908383623706636688843241160704
3 ==> -39148920061803210956691061036397438426365134774540146675772774556810648122275561412271743926625375084180283661376953125/2294287488095403763073974596800604256904498055825734116841519453845801891422853587965749144293625676308261780152348780234845939207145908383623706636688843241160704
4 ==> -24799801902139948285742077521010897923261300584495629346339410063333039539280530883995185222932342824043982837509521484375/4156317913216311164989084414493848291493655898235025573988259880155438209099372441966936855604394341138155398826718804773271628998452732579028454051972542103552
5 ==> -24799801902139948285742077521010897923261300584495629346339410063333039539280530883995185222932342824043982837509521484375/4156317913216311164989084414493848291493655898235025573988259880155438209099372441966936855604394341138155398826718804773271628998452732579028454051972542103552
6 ==> -39148920061803210956691061036397438426365134774540146675772774556810648122275561412271743926625375084180283661376953125/2294287488095403763073974596800604256904498055825734116841519453845801891422853587965749144293625676308261780152348780234845939207145908383623706636688843241160704
7 ==> -39148920061803210956691061036397438426365134774540146675772774556810648122275561412271743926625375084180283661376953125/2294287488095403763073974596800604256904498055825734116841519453845801891422853587965749144293625676308261780152348780234845939207145908383623706636688843241160704
8 ==> -39148920061803210956691061036397438426365134774540146675772774556810648122275561412271743926625375084180283661376953125/2294287488095403763073974596800604256904498055825734116841519453845801891422853587965749144293625676308261780152348780234845939207145908383623706636688843241160704
9 ==> 0
- : unit = ()

The correct result is zero.

(*) uname -a

Linux djinn 2.4.23 #6 Thu Dec 4 18:13:28 CET 2003 i686 GNU/Linux


I would like to know whether this can be confirmed.

-- 
regards,                  tf@cip.physik.uni-muenchen.de              (o_
Dr. Thomas Fischbacher -  http://www.cip.physik.uni-muenchen.de/~tf  //\
(lambda (n) ((lambda (p q r) (p p q r)) (lambda (g x y)              V_/_
(if (= x 0) y (g g (- x 1) (* x y)))) n 1))                     (Debian GNU)


===================