G++ issue (was "commentary")

[Mark S. Miller <mark>]

Date: Sun, 19 Nov 89 18:23:02 PST
   From: roger (Roger Gregory)


   Re: the g++ option

   Its not that clear cut, Michael Tieman is forming a company
   to support pd stuff, including g++ (which he wrote).

Don't you mean "... pd stuff, as well as g++ ..."? :-)

   Besides clearing up the library issue, this could get 
   g++ ported to the mac, Stallman & co are boycotting the
   mac,  Tieman won't be. 

Well we should certainly check with him.  However, on the Mac there
are also issues (of which I know very little. Heh?) like integration
with MPW, Object Pascal, MacApp, Mac compactable-handle objects, etc,
which was presumably why the 2.0 port took so long for Apple.  Even if
he is planning to do this, it would seem to be *way way* too risky to
count on it being done when we'd need it.  Also, to support Heh,
wouldn't we need it on the Mac almost immediately?

Also, what about the IBM PC?  Ravi would also need this in pretty
short order.  Having written a good code generator once, I remember it
being a lot of time consuming work.  I really think that compiling to
C is one of the things Bjarne did right.  As we say so often these
days, C is the world's portable assembly language.  As machine
architectures replace each other in rapid succession, I would bet that
any general sequential CPU will have a good C compiler before becoming
popular.  The only other languages I would say that for are those that
are built portably on a C base.

Nevertheless, I think we should talk to Tieman about all this.  And
soon.  Volunteers?

   One major reason to test g++, is to have something to 
   tweak Bjarne with.  Hugh Daniel is the logical person to 
   test g++ for this on the suns as he has installed 
   g++ on a lot of machines.

I took AT&T a *long* time to get 2.0 out the door.  I don't think we
should expect any new releases from them soon enough to affect us.  It
would still be nice to tweak Bjarne, but a low priority.  I do think
that having Hugh Daniel test g++ wrt both this issue, and our code in
general, would probably be a good idea.  Hugh, would this be a good
use of your time?  Are you interested?

P.S. The one thing I can think of that Tieman could have up his sleeve
that would change the above argument is an existing C code generator
for g++.  That'd be interesting.