Date: Fri, 27 Jan 95 17:35:25 CST
From: "Jason Hoffoss" <hoffo002 AT gold DOT tc DOT umn DOT edu>
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Major bug

Hi.  I've found a major bug that whoever fixes bugs might be interested 
in.  I've written a map editor for Doom, which I originally wrote to 
compile with Borland C++ v3.1.  Due to having way to little memory 
anymore, I had to do something, so I started porting it to DJGPP.  
Luckily, I know someone with experience with porting stuff from Borland to 
DJGPP, and he's helped a lot to just get it to compile and run.  Now I'm 
in the process of fixing up little things that don't quite work.  One bug 
turned into quite a major bughunt, setting testpoints in half a million 
points to track down just what the heck it had at various points along the 
way, and recompiling it comstantly and comparing it's results to those 
from the Borland C version.  But, I have finally found the problem.  What 
I had was something like this..

   typedef short int word;

   struct {
      char name[8];
      long unused1;
      word width;
      word height;
      long unused2;
      word p_size;
   } header;

   fread(&header, sizeof(header), 1, fp);

Then, I put in a debugging line like:

   printf("%d\n", sizeof(header));

Which gave me 24 for output.  If you look carefully, however, you will see 
that it should be 22.  Why is it giving be 24?  Have no clue, but it must 
be a bug.  I actually checked the sizeof() of each field in the header 
structure, and they all were fine.  Add up all six of these sizeof()'s and 
you get 22.  To fix my code, I just went and changed my old line to:

   fread(&header, 22, 1, fp);

Not really the recommented solution, but the only way to get the thing 
working again.  It works great again with it this way.  So, have fun 
locating that little bug.  :)

  -Jason