Front page | perl.perl5.porters |
Postings from October 2003
In message <rt-24091-65573.11.3042945319735@rt.perl.org> (on 2 October 2003
12:40:39 -0000), perlbug-followup@perl.org (jpl@research.att.com (via RT))
wrote:
># New Ticket Created by jpl@research.att.com
># Please include the string: [perl #24091]
># in the subject line of all future correspondence about this issue.
># <URL: http://rt.perl.org/rt2/Ticket/Display.html?id=24091 >
>
>The pow.t tests are somewhat surprising to me,
>knowing as little as I do about floating point.
>As the snippet below shows, the results are spot
>on for 17 through 27 *except* for 18 and 26 (for
>which they are very, very close to correct).
>
>ok 224 - -3 ** 17 got[-129140163] expect[-129140163]
>not ok 225 - -3 ** 18 got[387420489] expect[387420489]
># Failed at t/op/pow.t line 69
># got '387420489'
># expected '387420489' (within '0')
># 387420489 - 387420489 = -3.30872245021211069948563476828e-24
>ok 226 - -3 ** 19 got[-1162261467] expect[-1162261467]
>ok 227 - -3 ** 20 got[3486784401] expect[3486784401]
>ok 228 - -3 ** 21 got[-10460353203] expect[-10460353203]
>ok 229 - -3 ** 22 got[31381059609] expect[31381059609]
>ok 230 - -3 ** 23 got[-94143178827] expect[-94143178827]
>ok 231 - -3 ** 24 got[282429536481] expect[282429536481]
>ok 232 - -3 ** 25 got[-847288609443] expect[-847288609443]
>not ok 233 - -3 ** 26 got[2541865828329] expect[2541865828329]
># Failed at t/op/pow.t line 69
># got '2541865828329'
># expected '2541865828329' (within '0')
># 2541865828329 - 2541865828329 = -2.710505431213761085018632002175e-20
>ok 234 - -3 ** 27 got[-7625597484987] expect[-7625597484987]
>
>If it's of any help, adding -Dusemorebits does *not* help,
>you get exactly the same errors. Without the uselongdouble,
>the builds are flawless, with or without usemorebits. -- jpl
[...]
>Summary of my perl5 (revision 5.0 version 8 subversion 1) configuration:
> Platform:
> osname=irix, osvers=6.5, archname=sgi6-irix-ld
> uname='irix64 ent 6.5 07121148 ip27 '
IP27? R10000 chip, I'm guessing?
> config_args='-Duselongdouble'
> hint=recommended, useposix=true, d_sigaction=define
> usethreads=undef use5005threads=undef useithreads=undef usemultiplicity=undef
> useperlio=define d_sfio=undef uselargefiles=define usesocks=undef
> use64bitint=undef use64bitall=undef uselongdouble=define
> usemymalloc=n, bincompat5005=undef
> Compiler:
> cc='cc -n32', ccflags ='-D_BSD_TYPES -D_BSD_TIME -woff 1184,1552
> -DHAS_LDBL_SPRINTF_BUG -I/usr/local/include -DLANGUAGE_C',
-DHAS_LDBL_SPRINTF_BUG but not -DHAS_LDBL_SPRINTF_BUG_LESS1?
Interesting. I've seen this here also with an R5000 machine (non-64-bit),
but hadn't had time to check things out further. I've done some
rearrangements on the local smoketest configuration files and will see what
happens; my time to work on this is likely to be limited, however (lab
reports to grade this weekend and an exam to help grade next week, plus
TAing a lab, my research work, etcetera).
-Allen
--
Allen Smith http://cesario.rutgers.edu/easmith/
February 1, 2003 Space Shuttle Columbia
Ad Astra Per Aspera To The Stars Through Asperity