Date: Wed, 25 Nov 92 08:37:27 EST
From: DJ Delorie <dj AT ctron DOT ctron DOT com>
To: melvin AT math DOT psu DOT edu
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: disk crashes

>From: Richard Mansfield <melvin AT math DOT psu DOT edu>
>Date: Tue, 24 Nov 92 12:12:36 -0500

>I have been told by someone that his computer crashed and his
>hard disk was wiped out while running a program compiled by
>gcc. He thinks he traced the problem to a bug in cout.flush().

>This causes me some concern. Has anyone else encountered what
>they believe to be similar bugs? Is go32 inherently reliable?

>This note of course borders on rumor mongering and I am prepared
>to believe that my informant was done in by his own programming
>error and I am not about to delete gcc from my hard disk, yet I
>would appreciate further information.

This has been known to happen in a very small percent of the cases.
I've never had it happen to me, and I think I use gcc more than anyone
else.  Of course, I don't use cout.flush().  On the other hand, I've
seen other commercial packages (name withheld) with similar failure
rates, so I think that "inherent" reliability just doesn't apply to
software.  "Reasonable" reliability might be a better guage.

djgpp is distributed with no warrantee; any damage caused by it is
unfortunate and I (we?) try real hard to prevent it, but it sometimes
happens.  One of the readme's, I think, even mentions this.

If you find a bug that's potentially dangerous, please feel free to
fix it and post diffs.  The more people working on fixing bugs, the
less bugs there will be, and everyone will be happier.

DJ