From: Ian Miller <itmiller AT dra DOT hmg DOT gb>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Bug in `fstream' class
Date: 10 Feb 1997 13:54:02 +0000
Organization: Defence Research Agency
Lines: 36
Message-ID: <rkhgjkahs5.fsf@taz.dra.hmg.gb>
References: <x6raion8h7 DOT fsf AT dadzbog DOT ii DOT uni DOT wroc DOT pl>
NNTP-Posting-Host: 146.80.115.106
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

Jerzy DOT Witkowski AT ii DOT uni DOT wroc DOT pl (Jerzy Witkowski) writes:

> 
> A friend of mine found a bug in `fstream' class of the `iostream'
> library.  The bug makes impossible opening single file for both
> reading and writing.  It is present in both 2.00 and 2.01
> versions.
> 
> The example program is included below.  It produces the following
> results:
> on Sparc with Solaris 2.4 and gcc 2.7.2:
> 	[klm/k-m] (correct)
> on PC with DOS and Borland C++ 3.1:
> 	[mlk/m-k] (also correct, I think :-)
> on PC with DOS and DJGPP 2.xx:
> 	[   /   ] (not spaces, but (char)(255), also data file corrupted)
> 
> Any suggestions?

I have had problems with fstream too. I made them less bad by getting the
libstdcx sources and compiling them myself. Evidently, the distributed
binary library is not up to date (or wasn't the last time I looked).

For the record, my problem was that creating a fstream object with an omode
of ios::in|ios::out truncated an existing file to zero length. Once I had
regenerated the libstdcx.a library it stopped doing that.

I do still have one problem however. The get and put "streamposes" returned
by tellg() and tellp() are always the same. I think they should be
independent.

Any suggestions? ;-)

-- 
Ian Miller, Dorset, UK
DJGPP 2.01, Win95 DOS box (LFN undefined, FNCASE=y)