ohl@kivax.UUCP (Ohlhausen Abt.013.4) (02/08/90)
I found a bug in rcc (AT&T C-compiler in SCO Unix V.3.2/386) which occurs when
using floating point constants with significant digits after the decimal point:
float f;
f = 0.1;
printf("f = %f\n",f);
Output is: f = 1.000000
The bug is not in printf(), it's the assignment that goes wrong! Is there
already a fixed rcc out there? I can't believe that nobody found this
terrible error yet. I've read this group for quite a while but didn't see
any mention about this.
--
Ralf Ohlhausen In all probability it's my opinion only!
Mannesmann Kienzle GmbH ohl@kivax.uucp
7730 VS-Villingen Europe: ..!mcvax!unido!kivax!ohl
FRG USA: ..!uunet!unido!kivax!ohlwsinpdb@eutws1.win.tue.nl (Paul de Bra) (02/09/90)
In article <590@kivax.UUCP> ohl@kivax.UUCP (Ohlhausen Abt.013.4) writes: >I found a bug in rcc (AT&T C-compiler in SCO Unix V.3.2/386) which occurs when >using floating point constants with significant digits after the decimal point: > float f; > f = 0.1; > printf("f = %f\n",f); >Output is: f = 1.000000 This is an SCO feature :-) as it does not occur with the AT&T C-compiler in AT&T Unix V.3.2/386. Or else it is a feature of the 80387 emulator if you don't have the coprocessor (you didn't say, but you should, whenever you mention a floating point problem, as the emulator is known to be buggy). Paul. (debra@research.att.com)
ohl@kivax.UUCP (Ralf Ohlhausen) (02/12/90)
In article <590@kivax.UUCP>, ohl@kivax.UUCP (Ohlhausen Abt.013.4) writes: > I found a bug in rcc (AT&T C-compiler in SCO Unix V.3.2/386) which occurs when > using floating point constants with significant digits after the decimal point: > > float f; > f = 0.1; > printf("f = %f\n",f); > > Output is: f = 1.000000 In the meantime I found out that the bug occurs only if the environment variable LANG is set to german_germany.8859 during the compilation. Poor German :-( . BTW: Greetings to the french_france.8859 people. You have the same problem! :-) Since the assembler code - produced with the -S option - shows no difference, I believe that he's the one who doesn't like my language. -- -- Ralf Ohlhausen * e-mail: ohl@kivax.uucp * In all probability Mannesmann Kienzle GmbH * Europe: ..!mcvax!unido!kivax!ohl * this is once again 7730 VS-Villingen FRG * USA: ..!uunet!unido!kivax!ohl * only my opinion !