Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site rayssd.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!rayssd!dhb
From: dhb@rayssd.UUCP
Newsgroups: net.games.rogue
Subject: Re: The FLAME hits --More--
Message-ID: <556@rayssd.UUCP>
Date: Tue, 6-Nov-84 12:36:49 EST
Article-I.D.: rayssd.556
Posted: Tue Nov  6 12:36:49 1984
Date-Received: Thu, 8-Nov-84 00:26:36 EST
References: <237@wjvax.UUCP>
Organization: Raytheon Co., Portsmouth RI
Lines: 26

I think that the save game bug is one of the worst things in the entire
game.  Most of the rest of the things can be worked around by careful
playing but not this one.  If you have ever seen the source to the game
you will understand not only the cause of this bug but the complexity
of attempting to resolve it.  The way the program saves the game is by
finding out the beginning address of the data space, doing an sbrk() to
find the end of the data space, and then writing the entire data space
to the file.  The restore routine simply finds out the size of the file,
sets the size of the data space to that amount, and then reads the file
in.  When I first saw this I almost choked.  This is not to say that I
think the authors are poor programmers, just that the code is EXTREMELY
un-portable (I was trying to port an old version (3.8) of the program
to another operating system at the time!).  This method of doing the
save/restore is also highly prone to problems (witness the current bug)
due to the fact that a large portion of the data space contains pointers.

Just let me state again that I am not critizing the quality of the
save/restore routines coding, only it's portablity.  If any of the
authors are interested, I have a save/restore routine about 95%
finished which should alleviate these problems.
-- 
	Dave Brierley
	Raytheon Co.; Portsmouth RI; (401)-847-8000 x4073
	...!decvax!brunix!rayssd!dhb
	...!allegra!rayssd!dhb
	...!linus!rayssd!dhb