Re: Writing a Mac Ada compiler

[jmj@redstone.army.mil (Mark Johannes)]

In article <3ji8rc$rj8@nic.umass.edu>, cons116@twain.oit.umass.edu (Mike White) writes:
>Arthur Evans Jr (evans@evans.pgh.pa.us) wrote:
>
>: If you have time on your hands and want to address this problem, port
>: GCC 2.6.3 to the Mac, and then bootstrap GNAT.  That's a much more
>: modest task and will result in a first class compiler.
>
Amen...

>Would it really be that much easier to port it? Have you seen the source
>code?  I haven't, so I am wondering how well designed/documented it is.
>I realize it's a pretty big task, but 10+ man-years?  O.K.  I'll admit
>it's probably not a one-man project, but if others were interested in
>working on such a project...
>

Yes you would have to learn something about the internals of GCC but
you would not have to write the large mass of code known as a compiler.
Unless I am horribly off base here, your effort would focus on providing
host configuration files and host specific code.  Since both the m68k and 
PowerPC families are well covered by GCC/GNAT.   

>  I mean seriously, I wrote a simple lexer and parser (both using Ada
>by the way) in my systems class both in a semester.  Both of these
>were easily extensible for large projects, and the grammar
>for Ada is widely available.  Beyond that, converting to machine code
>would seem my biggest problem, never having done mac or powermac assembly.
>Interfacing with mac system calls seems pretty basic since they are well
>defined.  I imagine one would convert the parameter conventions from Ada
>to the system via MacHeaders.

yes a lexer and parser may not be too tough for you to develop but.. you need
2 code generators, linkers, assemblers, debuggers, and Ada runtime support
to have a full environment.  Again your effort should be restricted to providing
MAC specific support. -- not reinventing the wheel.

I personally don't relish the idea of even using the machine_code package
much less implementing it.

>  Just some thoughts.  I need to do some more research obviously before
>making any attempt.  Anyone interested in participating on a joint effort,
>either working from scratch or doing a port?  Please reply here so that
>everyone interested can read it, myself included.
>
>mike white

And let's not forget this issue: support and maintenance.  The GNU tools
have a lot of users.  There is a large body of code shared between the C, Ada,
Pascal, and now Fortran users.  This significantly increases the user population
for the toolset.  Yes the port itself may only have a handful of users..but the 
code generator and other support components are shared and have who knows
how many users.  So the odds of finding bugs at a serious level are significantly
reduced.

Finally, a port of GNU Ada gives a lot more to the MAC community than an Ada
compiler.  It lays (all??) of the groundwork for ports of other GNU languages and
gets enough tools in place to allow ports of other GNU tools.  Wouldn't GNU C
work when/before GNAT did?  C++, Objective-C, Fortran, and Pascal would not
be far behind.   it starts as a trickle of tools and ends as the Hurd .. (sorry could
not resist).

I probably missed something... but you get the idea.

Joel Sherrill
jsherril@merlin.gcs.redstone.army.mil