Newsgroups: comp.os.msdos.djgpp
Subject: Re: files not closed after GPF
From: richard AT stardate DOT ca (Richard Sanders)
X-Newsreader: WinVN 0.99.9 (Released Version) (x86 32bit)
References: <20010805030625 DOT 08871 DOT 00002852 AT ng-fr1 DOT aol DOT com> <9kjsnd$ian$1 AT newsg3 DOT svr DOT pol DOT co DOT uk>
MIME-Version: 1.0
Content-Type: Text/Plain; charset=US-ASCII
NNTP-Posting-Host: wlp48.rapidnet.net
Message-ID: <3b6dc59f_1@rapidnet.net>
Date: 5 Aug 2001 15:15:59 -0700
X-Trace: binaries.vphos.net 997049759 wlp48.rapidnet.net (5 Aug 2001 15:15:59 -0700)
Organization: Rapidnet Internet
Lines: 24
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

In article <9kjsnd$ian$1 AT newsg3 DOT svr DOT pol DOT co DOT uk>, 
matt AT the-good-stuff DOT freeserve DOT co DOT uk says...
>
>
>>
>> Is it a bug ?
>
>I think you are doing about as well as can be expected, only losing
>one line line that.
>crashing with files open can actually wipe your whole file system if
>you are exceptionally unlucky.
>
>If your program was opening the file itself, it could adopt an
>"open,append,close quickly" strategy but redirecting the output makes
>your shell handle this.
>
>You could try outputting a few line feeds after critical information
>to flush the shell's buffer to disk.

Better yet flush the file to disk after every write. If you are using buffered 
file io (fopen), use fflush( FILE *foo );, for unbuffered io (open) use 
_dos_commit( int foo );.  The flushing will be done before the crash and the 
info should be in the file, baring other problems.