From: denlee AT see DOT sig (Dennis Lee)
Newsgroups: comp.os.msdos.djgpp
Subject: (second try) Possible bug in startup code for math emulator
Date: Wed, 29 Oct 1997 06:08:25 GMT
Organization: INTERNET AMERICA
Lines: 32
Message-ID: <94C27C3C43AA5A64.AD775D4D7DFCE1BB.137E2C37FABAE847@library-proxy.airnews.net>
NNTP-Proxy-Relay: library.airnews.net
NNTP-Posting-Time: Wed Oct 29 00:01:55 1997
NNTP-Posting-Host: 209-20-1-79.dialin.interlog.com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Hello again,

I didn't receive a reply to my earlier post, so I'm trying again 
from a different news server in case the last one didn't 
'spread'.

My program which uses floating-point math fails
in a Win95 DOS Shell when trying to use math
emulation (387=N).  Under a real DOS session,
it runs fine.  I was able to track down the problem,
and apparently any program that uses floating-point
will fail under the above conditions.

A specific example is:

#include <math.h>
int main() {
    float x = floor(-1.5);
    return 0;
}

This has been tested on a Pentium and PPro.
Is this a known problem ?

Thank you for any help.

--
Dennis Lee       E-mail: denlee AT ecf DOT utoronto DOT ca
Waveform Archiver -- Lossless Audio Compression
DL1 & DL3 2-Pass Color Quantizers (Source Code)
Popular CPU/FPU/Memory x86 Benchmarks
www.ecf.utoronto.ca/~denlee/software.html