Calculations incorrect with large numbers

[Xhacker Liu]

Binary package hint: gcalctool

When meets big numbers, Gcaltool 5.29.92 calculates wrong answer.
For example, "888!":
When I use python, it returns 257365811240746258261058312635496048130210642911052728912405050489847845936124604455726245537138568519919138869993820540619317156386110281318043523819271133040245050833530381384223242463725798006902411731729593770548785690354781086905842723308884801601200379052751917433951594795069246539623889566172402612394273813442913138634455511343892185394277862364138060724692422176326391957764020447386941437446725478761545107390817405634196802572153899519689487073825003215226617854722625058758294222344083356593753257086354039728300782461317497535062225687078681424411893621525850551269343434971126659954733715916782958472453656768810440701195867642664276246868388275821267618720767840980999383693936888759532254366317019814634072835535528238339276696529590492607088777154869458497157498973960649754315551943201432369267509169588502957749960541697921095814304037256642324456938413254625533153280856940717899783245778945051168906015433910286854807561451501938458959243234283134088179776468080393516478163145085578656974388367221108095677627567015952254146171246437988038149150987707167315073537790062736939314320210458720139105535344696345447770657457362031933212587185848613853398987903918414682644458052794592851305410955768885474260613862805460079057807727238845021832356538307021541713160320875576539401590159447926998635089309190781896462345990784880322764271663904123377742519178473040194012177227689415411742481696796207548195180131422419383028352271214484900120165291146804072489789248793346781239796284546531818941890571677176899236847462415968281683926928007538265304095422178343439306412724651705340187355911592282977060598094755781754218900772740050037966772213112296704368289348866131536557227203591833504468255928529993693686570881138553294753614798339407527326181492069008868558254336122129801882121115421377833709324088258996044105013303675033524073942148782155553588617920517036672360857955988473054416496830250418320295432036884294666305652837013806869131833276398423114995974063714689045182764541217341440000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000.
But gcaltool returns 25736581124074625826105831263549604813021064291105272891240505048984784593612460445572624553713856851991913886999382054061931715638611028131804352381927113304024505083353038138422324246372579800690241173172959377053910000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000.

ProblemType: Bug
Architecture: i386
Date: Wed Mar 3 13:25:19 2010
DistroRelease: Ubuntu 9.10
ExecutablePath: /usr/bin/gcalctool
InstallationMedia: Ubuntu 9.10 "Karmic Koala" - Release i386 (20091028.2)
NonfreeKernelModules: nvidia
Package: gcalctool 5.28.2-0ubuntu2
ProcEnviron:
 LANGUAGE=zh_CN.UTF-8
 LANG=zh_CN.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.31-20.57-generic
SourcePackage: gcalctool
Uname: Linux 2.6.31-20-generic i686

[Xhacker Liu (xhacker)]

Not only 888!, other big numbers can cause the bug too.

[Robert Ancell (robert-ancell)]

Could you give some other test cases? Thanks!

[Xhacker Liu (xhacker)]

88888^100:

Gcalctool:
766149387302276358647668963491427744310804798315976240953970499821711231191237430334433556085573811714287674335578790498107987999820495052364976767206745507081913259956629403143506051971715909926796633437903308858976400000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000

bc:
766149387302276358647668963491427744310804798315976240953970499821711231191237430334433556085573811714287674335578790498107987999820495052364976767206745507081913259956629403143506051971715909926796633437903308858978699832696595344598777484807975307251380452083847559295720545304745334889037935791289779623465801577577693300966123018918250337902527522498468494897254033434267936929202727474649572488672832848517655542855762115946055745019554297960959548960911915572020601859582331181601050853376

[Oliver Joos (oliver-joos)]

The test case in comment #4 is not wrong but only imprecise - numbers of digits are equal: 495.

But the description of this bug shows a severe problem with the new display routines. If the Output format is "Scientific" or "Engineering" then the display shows "2.573658112×10²²³⁴" or "257.365811241×10²²³²" which is correct. With "Decimal" I confirm it to be wrong. And if I switch to "Binary" a core is dumped!

gcalctool 5.28.x did not have this problem. It switches automatically to Scientific format if there are more than 100 digits - even when in Binary mode: 1.01110111e+1110011111110

Hint for bug hunters: with Decimal format gcalctool clips the result of 888! to 512 chars. ;-)

[Robert Ancell (robert-ancell)]

This wrong result is due to memory corruption as can be show with valgrind. This issue was still present in 5.28.

[Enrico B. Accenti (eaccenti)]

Same behaviour on 5.31.2 . I found it out summing hundreds of '9' to 1.

[Xhacker Liu (xhacker)]

Why this serious bug is still not fixed?

[Enrico B. Accenti (eaccenti)]

Seems fixed on 6.0.1.
 Tested on
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=11.10
DISTRIB_CODENAME=oneiric
DISTRIB_DESCRIPTION="Ubuntu oneiric (development branch)"

[Xhacker Liu (xhacker)]

@Enrico B. Accenti
I tried 6.0.1 just now, but unluckily the issue is still exist. (I tried "888!" test case)