Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!uakari.primate.wisc.edu!csd4.csd.uwm.edu!uxc.cso.uiuc.edu!tank!uwvax!puff!rt18.cs.wisc.edu!blochowi
From: blochowi@rt18.cs.wisc.edu (Jason Blochowiak)
Newsgroups: comp.sys.apple
Subject: Re: C development environment for IIGS
Keywords: C,IIGS,development
Message-ID: <2940@puff.cs.wisc.edu>
Date: 8 Sep 89 02:09:06 GMT
References: <197@fjcp60.GOV>
Sender: news@puff.cs.wisc.edu
Reply-To: blochowi@rt18.cs.wisc.edu (Jason Blochowiak)
Organization: U of Wisconsin CS Dept
Lines: 80

In article <197@fjcp60.GOV> brinkema@fjcnet.GOV (John R. Brinkema) writes:
>I am about to purchase a C development enviroment (C compiler, plus
>friends) to do some software development on my IIGS. I am interested
>in what the experienced developers recommend.
>
>I am a very experienced C programmer (a lot under UNIX, some on a
>Mac using THINK's Lightspeed C .. which I like a lot!). Of course I
>have a IIGS, a hard disk, 1.25 M memory, but little else other than
>a great deal of interest.
>                                          tnx. jb

	Well, don't get your hopes up too much... Both of the C compilers
available for the //gs are mediocre at best. There's APW C, which seems to
be a port of Megamax C with a 65816 code generator stuck on the end, and
Orca/C which (as far as I know) is a complete from-scratch compiler.

	Both are buggy. According to a friend (I haven't checked it yet),
if you do "a += 1.0" where a is a float, APW C will generate code that
crashes. Orca/C is worse - on every program that I've written, I've
discovered a new bug - the worst is that they're so basic, I'm frightened
about how many subtle bugs will be generated. Two blatant examples: In one
program, I do a switch (char variable) with a few cases and a default. The
compiler either 1) Didn't generate the code associated with the default
clause (what, a parse tree with dutch elm disease? Look, ma, another branch
just fell off...), or 2) Just didn't generate the proper branch. I didn't
poke that far into the disassembly (just far enough to figure out that it
was the compiler, and not my code or my imagination). The second example:

typedef	struct { int junk } x;

	x bug[] = {{1},{2},{3}};

main()
{
	int i;

	for (i = 0;i < 3;i++)
		printf("%d ",bug[i]);
}

	The program will print out "1 1 1 ". What happened was at some
point the compiler decided that (sizeof(x) == 0). If the array is dimensioned,
e.g. "x bug[3] = ..." the code works as it should.

	The Orca/C libraries are much smaller than APW C's - I haven't had
much reason to check if they work. printf(), gets(), and atof() appear to
work as advertised... By smaller, I mean the code is smaller - they appear
to have the complete ANSI C library routines.

	Neither of them is particularly speedy, but with a TransWarp GS,
they're tolerable (without one, though, you'll develop new "waiting for the
compiler" habits). I think that Orca/C is faster, but haven't done any
testing.

	Orca/C comes with a (buggy to the point of uselessness) desktop
environment, as well as an (fairly bug free) APW compatible text-based
shell (the desktop environment also allows access to the shell). APW C
requires the APW shell (which also includes a pretty nice macro assembler
and some other utilities). The desktop environment includes a (Pascal
oriented, although not much) source level debugger, in which there are
also bugs (either that, or Orca/C creates bad information for it). There's
a text based debugger for APW (when I say APW, I'm referring to the APW
shell as well as the text-based Orca shell) that's pretty nice, considering
its simplicity (it's not even symbolic).

	Orca/C allows you to create NDAs and CDAs as well as normal programs
(EXE & S16). APW C requires a bunch of weird shit with setjmp to support
NDAs, and can't be used to write CDAs. Orca/C also supports more calling
conventions, which make doing certain things more convenient.

	Basically, Orca/C has more "stuff", but is significantly buggier
than APW C. APW C is also not ANSI compatible (no function prototyping...).
Oh, speaking of function prototyping, Orca/C doesn't allow prototyping
toolbox calls, which is really annoying, considering the large number of
calls in a standard desktop program. Neither is great, but they work, sort
of...

	If you (or somebody else) has more questions, ask away...

	Jason