8 Classes can see "my-field" (which is correct), but the codegen is a field access, when it should be a send to the "Main" proto. "emit-c-call:builder:" methods of FieldGetter and FieldSetter need to be fixed. [Fixed 2007.9.1]
10 To optimize field access, "MethodContext lookup-function:" would need to explicitly check fields before checking for function calls on "this". But that's not semantically safe, as a subclass could override the getter and/or setter.
17 Can't use setjmp()/longjmp() to pass the exception object, as they only pass an "int", not a "long". This is easily fixed by only passing 1 there, and using a global to hold the exception value. That's not thread-safe, but our exception chain isn't thread-safe currently either.
19 ClassInfo stores size in bytes, rt number of fields. That doesn't work when the generated C code is moved between 32 and 64 bit platforms.
21 Int should go to 64-bit on 64-bit platforms... but that probably won't be necessary to make it work.
29 Current crash is due to a problem with precedence. Need to separate assignments from other binops:
37 The old parser doesn't have that problem, but at the expense of not being clean about its syntax. It ought to be fixed before the Trylon 2 release (assuming we don't switch to the new parser).
55 unary+, unary-, ~ [Used to have !.]
71 Field access optimization. Strictly optional, since even in constructors we can't be sure we aren't dealing with a subclass that overrides the getter and/or setter. But it might be a helpful optimization for those who know what they're doing.