From: Mark Hull-Richter <mhr AT sparc DOT SanDiegoCA DOT ATTGIS DOT COM>
Subject: Odd behavior - summary
To: djgpp AT sun DOT soe DOT clarkson DOT edu (DJGPP Mailing List)
Date: Thu, 2 Mar 1995 14:33:45 -0800 (PST)

FYI, as promised, here is a summary of the advice I received.

The most prominent suggestion was to compile with both -g and -O.  I did
that.  Now a seed which bombed the -O (only) version right away takes as many
as five rounds before it blows up.  The most recent error that occurs shows up
in edebug32 as an exception 12 (0xc), error code 0x20000.  I don't know what
that means, but the register dump shows a destroyed %ebp.  This seems to
happen when a particular selection is set up that loops twice and bombs after
the second iteration.  This is interesting, as the function in question is
executed dozens of times before the error occurs, but I'm still on it.

Another suggestion, re the program sizes, was to use strip - I will if this
continues too long....

There was at least one suggestion to use -S and look at the assembly language
of the offending module.  I may do that if I have to, but a) the module is a
bit large for that and, more importantly, b) I'm no expert at assembler (yet
:-).

One person suggested I use the errors in BCC to track down incompatibilities
and fix the program. I sent that person a response, but on re-reading would
like to add that I did try that.  The problem is that the program dies under
BCC during primary initialization, where it is trying to allocate some data
structures.  This happens every time, from the command line or under the IDE,
and it blows the machine away every time.  I tried to sneak up on it in the
debugger, but I could never get it to stop in the right place, and as soon as
it went that one or two lines too far, blam.  Also, I kept seeing anomalies in
the generated code that led me to believe that the problem (there) was with
BCC, not my program.  TCC had the same type of difficulty, except that it
couldn't seem to accumulate into a variable and would always wind up
attempting to allocate a 0 byte datum, and that killed the machine as well.  I
have little enough time to work on this as it is, so I decided to follow the
path of least disruption, which is so far DJ's gcc.  It works most of the
time, and I should be able to isolate the bad case where it keeps failing.

New question: what are the exception codes / unsupported INTs and where can I
look up them and their associated errors codes?  I tried using info, but I
guess the interface either doesn't have the answer, is missing a needed file
(which one?) or is too different for me to follow (yet).

I haven't given up.  Thanks for all the input.  I'm still on it.

-- 
 ++==     AT&T   Mark A. Hull-Richter, Consulting Engineer  (310)524-5782 voice
+GIS==   Global  Teradata Decision Enabling Systems Center   427-5782 VoicePlus
=++=== Information  100 North Sepulveda Boulevard, #17-259  (310)524-5517 FAX
 ====   Solutions   El Segundo, CA  90245       mhr AT sparc DOT SanDiegoCA DOT ATTGIS DOT com