Re: proposed patch for PR 72 (method lookup in interfaces)

[Per Bothner <per at bothner dot com>]

Godmar Back <gback@cs.utah.edu> writes:

> The whole idea of using one union 'tree' to represent everything in
> the system seems like it came from a bunch of lisp programmers 
> with no appreciation for types writing their first C program.  

Yes, in retrospect having a 'union tree' for everything was probably
a mistake.  It does have some advantages.  Still, if you used C++
and inheritance (not an option, alas), you get the flexibility of
a common tree type, but can use subtypes to get better typechecking.

> But why, for instance, did I have to say CLASS_INTERFACE(TYPE_NAME(clazz))
> to determine whether "tree clazz" is an interface?  Why wouldn't 
> CLASS_INTERFACE(clazz) work?  If types were used in a meaningful way, 
> the compiler would have warned me.  The compiler can be your friend.

Well, you can compile gcc with -DENABLE_CHECKING (which is set when
you configure with --enable-checking) to get run-time checking.
(Run-time here means when gcc is run.)  This feature was added
relatively recently (in the last couple of years).
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/