Re: Adding zlib to GCC

[Joe Buck <jbuck at racerx dot synopsys dot com>]

> It's certainly difficult to *release* front-end changes because of
> back-end and other changes.  People are talking about adding Ada and
> Pascal to GCC.  I've heard talk about C#.  If all that happens, and
> attendant libraries, I think it will be hard to keep the GCC release
> cycle moving.  If Java/Ada/C++ needs to spend 6 months retooling,
> which might be perfectly reasonable, it seems like a shame to make
> everything else wait for that.

One way some software companies handle things like this is with a "train
model".  Think of the core of the software as a train.  There is a
schedule; the train is going to leave the station at some time.  Some
components of the train are critical; if they are late, we have to stop
the train.  Other components are not critical; if they are ready and
integrated in time and considered worthwhile, we hook them onto the train.
If not, they wait for the next train.

Since a new, stable C++ ABI is a primary goal of gcc 3.0, the C++ front
end is a critical component.  However, Pascal is clearly just a
nice-to-have; it can ride the train if it is ready but it can't hold it
up.

If we're ever going to get 3.0 out, we have to have a very rigorous
definition of what's critical and what's not.  The non-critical stuff
can go in IFF people who really care about it get the work done and
have it ready with minimal cycles from folks like Mark who have critical
responsibility for the whole thing.