From kde-devel Sun Aug 08 12:06:43 1999 From: Uwe Thiem Date: Sun, 08 Aug 1999 12:06:43 +0000 To: kde-devel Subject: dump question X-MARC-Message: https://marc.info/?l=kde-devel&m=93411400314046 I'm stuck with a little problem I'm too dump to solve it seems. Maybe someone else can shed some light on it. This is the code: double tmp = mRealCenter - mRealWidth / 2.0; QString s; s.setNum( tmp, 'f', 60 ); cerr << s << "\n"; for ( i = 0, p = mReals; i < 3; i++, p++ ) { s.setNum( (double) i * mRealWidth / (double) mWidth, 'f', 60 ); cerr << s << "\n"; s.setNum( tmp + (double) i * mRealWidth / (double) mWidth, 'f', 60 ); cerr << s << "\n"; } cerr << "\n"; And here comes the output: 0.291609459012032723457252814114326611161231994628906250000000 0.000000000000000000000000000000000000000000000000000000000000 0.291609459012032723457252814114326611161231994628906250000000 0.000000000000000000733384188652038425800936390023343002917542 0.291609459012032723457252814114326611161231994628906250000000 0.000000000000000001466768377304076851601872780046686005835084 0.291609459012032723457252814114326611161231994628906250000000 The first line represents the variable "tmp" in the code. Then follow three pairs of lines. The first represents a small number calculated in the code and the second the sum of "tmp" and this small number. As you can see the small number is about 19 levels of magnitude smaller than "tmp" but still significant. What baffles me is that although the small numbers _are_ significant and different the sums are all equal. Sure, if the small numbers get much smaller this problem will occur eventually. But why so early? At this stage the sums should differ but don't. #-( Uwe ------------------------------------------------------------------------- Uwe Thiem Tel: +264 - 061 - 244511 P.O.Box 30955 Fax: +264 - 061 - 244511 Windhoek Email: uwe@uwix.alt.na Republic of Namibia uwe@kde.org http://www.kde.org ********************************** You can still escape from the GATES of hell: Use KDE! -------------------------------------------------------------------------