Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!umd5!uvaarpa!mcnc!duke!jwg
From: jwg@duke.cs.duke.edu (Jeffrey William Gillette)
Newsgroups: comp.sys.ibm.pc
Subject: When will Microsoft give us a *REAL WORLD* C Compiler
Message-ID: <11622@duke.cs.duke.edu>
Date: 21 Apr 88 01:03:02 GMT
Organization: Humanities Computing Center, Duke University
Lines: 47

Last year I was working on a large project with Microsoft C 4.0 
in the large memory mode.  Half-way through, I noticed that 
uninitialized global variables (the kind put in the BSS segment) 
were being mysteriously trashed at random intervals.  A few weeks 
with CodeView, and I discovered that malloc() was allocating 
storage right over the top of uninitialized globals!  A tar to 
Microsoft (then I had access to Dial) brought the answer that (I 
paraphrase), "We know there is a problem with uninitialized 
globals, others have told us of the problem, we will fix it with 
the next version of MS C." 
 
This year I am working on a smaller project for an independent 
client.  I develop a program using the large memory model and MS 
C 5.0.  All of a sudden, some uninitialized globals start being 
trashed at random intervals.  I reach (by instinct) for CodeView, 
set a few watchpoints, and discover (surprise, surprise) that 
malloc() is allocating storage right over the top of 
uninitialized global variables!
 
Probably a dozen UseNet aces have already pointed up the same 
problem.  And, this may not be a problem for programs tht keep 
all their data in the DGROUP segment (I haven't done much with 
uninitialized data in small or medium models).
 
My question is simply, when is Microsoft going to get the message 
that real world programmers don't use the small memory model all 
the time.  Sure the medium model (a la Quick C) provides for far 
code and near data - give me a break!  The opposite would have 
been much more useful!  Now I no longer have access to Dial, and 
can no longer get the "official" Microsoft position on such 
subjects.  Anyone want to venture a guess as to when Microsoft 
will get on the ball and give us a *REAL WORLD* C compiler that 
handles both initialized (DGROUP) and uninitialized (BSS) global 
data in a large data memory model?  Until then, my perception is 
that one uses uninitialized global structures (in large data 
models) at one's own risk!
 
Jeffrey William Gillette
 
Jeffrey William Gillette	uucp:   jwg at duke.edu
Duke University                 bitnet: DYBBUK at TUCCVM




-- 
Jeffrey William Gillette	uucp:   jwg at duke.edu
Duke University                 bitnet: DYBBUK at TUCCVM