Conference 7.286::atarist

    There's differences in price because they're all different products!
    All with pluses and minuses...
    
    If you've got enough memory on your machine you might consider starting
    with sozobon C (free!!!) then buying something more substantial once
    you know exactly what you want from the language...

Sozobon C + Gulam shell + Microemacs 3.10_j isn't bad for the price.  I
have Sozobon C and Microemacs here on PRNSYS.  You wouldn't want this if
you needed floating point, but there is a new release of Sozobon coming
out soon, according to what I heard on GEnie.

Jerry Feldman has tried all of these compilers.  Perhaps he'll pop in and
say something.

    Since Jeff has prompted me.
    First, Sozobon C is a good start.
    I used Lattice for developing part of ChessMaster 2000. I had all sorts
    of problems. I do not recommend Lattice. (However, to be fair, it was
    back at release 3.x).
    
    Prospero C is compliant with the ANSI Standard. It has a good
    workbench, a debugger, and reasonable documentation and GEM library.
    
    Mark Williams C is the C supplied as part of the Atari Developer's Kit.
    It comes with a command line Unix style shell. It is a very credible
    compiler for the professional developer.
    
    My preference is Megamax Laser C. While I think that MWC generates
    slightly better code, I prefer the Laser Shell which includes a Cache,
    editor, Disk Utility menu, ... (Both Laser and MWC fully support GEM,
    but running GEM programs from MWC is somewhat awkward).
    
    Many developers are switching to Borland Turbo C. Turbo C is not yet
    officially available in the US. (I don't know what the status is in the
    UK). I have checked the PC version out, and it is excellent.
    
    ---- The bottom line is, try Sozobon. If you need a fully supported
    compiler, then try one of the above.
    
    

Here's another approach - as far as I know, most of us that are
participating in this conference and are activly developing code for the
ST are using Mark Williams C.  That fact would make me lean toward MWC,
because I could ask questions here and get an answer.  Be aware, tho:

	- MW's machine level debugger doesn't work on TOS 1.4
	- MW's RAM disk doesn't auto-boot on TOS 1.6
	- MW has no plans to fix either of these
	- Nobody at MW understands the ST (they all left about two years ago.) 

Last I knew, developers in the US were still getting Alcyon C.  That's
what I got last year when I registered.  MWC is available to registered
developers at a discount.

    > MW's RAM disk doesn't auto-boot on TOS 1.6
	MWs RAM disk does not work on my 1MEG (old) 520ST either.
    > Re: Alcyon C.
    I do not have the most recent version of Alcyon C, but when I did
    Chessmaster 2000, Alcyon was also unusable. It generated incorrect
    object code for large modules.I ended up having to go back to an
    Updated Lattice which generated very inefficient code.
    
    I also agree with jeff. I think that MWC is probably the best approach
    if you are going to get help on this conference, and MWC does generate
    good code. I do not remember the numbers, but MWC does better in the
    benchmarks than the others.
    
    

I've never had a problem with MW's RAM disk on TOS 1.0 or 1.4, either
bootable or not.  I wonder what went wrong in your configuration.   Were
you perhaps using an SH204 hard drive?  There was some bad interaction
with the SH204 and the early AHDI software that required that you use a
very high drive letter for the RAM disk, such as 'H' or 'K'.

    A C-Compiler?
    GOD Help...
    If you are starting, then why not use a REAL language. i.e. Modula2?
    Klaus (who stopped needing a debugger...)
    

    
    Hello,
    
    Another approach,
    
    Hisoft in th UK are offering a cheap beginners C which is INTERPRETED,
    i.e. it behaves like BASIC.
    
    This means you can try a code construction in direct mode, and get
    immediate error messages, or you can type in a segment and run it from
    the editor, without a compile  link stage. OBVIOUSLY it aint going to
    be fast but it must be a lot easier for learning! 
    
    Hisoft say that when you have written your proggy and it runs under the
    interpreter, then you can use a "standard" C to compile it for speed or
    compactness or whatever your particular bag is!
    
    Regards
    
    Pat K.

    Modula-2 ?
    GOD Help...
    Why not use a REAL language. i.e. straight 68000 machine code??
    Jamie (who, after adopting this approach stopped needing a brain!!!)

  I suppose I might as well lay my cards on the table.  When I first bought
my ST, I had bought Lattice C and used that for some time, but was never
really pleased with it, especially the linker, it took forever to link stuff.
I then got MWC.  I liked this compiler because it provided a command line
environment.  In the past 6 months, I have tried the GNU C compiler.  I have
one complaint with the MWC and GNU compilers, and that is their assemblers
don't use standard M68000 mnemonics for the instructions.  Particularly in
the notation for the various addressing modes.  Since I develop M68000 code
at Digital, it is a pain.  Here is how I would rank the three C compilers I
have used.

1) GNU C - Pros - This is an excellent compiler, which generates good code.
   It's free.  Is command line driven (ie doesn't use a gem shell).  Is probably
   the most ansi compliant you can get. 
   Cons - Takes lots of memory to use.  Pretty much 2megs ram needed, 4 is
   recomended.  The assembler doesn't use standard 68000 notation.

2) MWC - Pros - Generates good code.  Is command line driven.
   Cons - Pretty much no longer supported by Mark Williams.  Doesn't use
   standard 68000 mnemonics.  Isn't ansi compliant.

3) Lattice C* - Pros - Can't think of any.
   Cons - Wasn't ansi compliant.  I don't think anyone is supporting it.  Was
   pretty buggy in the code it generated.  The library wasn't complete.  Very
   slow to compile/link.  Used a gem shell instead of command line driven.

* Note - This was version 3.? I think there was an update to the compiler, but
  I never got it to see if it was any better.

Bill Powers

    I don't remember all the details. I do not have a SH204. After I had
    trouble, I went back to use your cookbook. Still had a problem. I think
    that the only way I was able to get the MW ram drive working was to
    first start SI RAM and then stop it. The MW RAM will work. Note that I
    normally do not use SI RAM. If you are interested in pursuing it off
    line, I can play with it a bit, and send you the details.
    

Oh, I remember this conversation now.

    I believe that the problem with my MW RAM disk has to do with the fact
    that I have an ADAPTEC host adaptor. I think it will work if I unplug
    the disk.