[PATCH] New pragma handling mechanism.

[Dmitry Kurochkin]

> What _I_ want appears to be much more like what Dmitry wants,
> but Zack doesn't.  I wanna see _these_ pragmas totally inline
> with the text of the program, so that I can just treat them
> as any other text that needs parsing.  Then my "#pragma omp for"
> parser does its thing and recurses on c_parser_for_statement,
> validates the restricted expression forms, and modifies the IL
> to be whatever we decide the right thing is.
> 
> I was kinda happy with Zack's "end-pragma-and-continue" thing
> becuase it will still let me do that, only with a small jog in
> the middle.
Yes. I thought that "difficult" pragmas, such as OpenMP pragmas, will
have no handler, and will be parsed by the front end internally. This
will require some common code for such pragmas in each front end. But
I think Zack's approach introduces another problem, we can't use front
end specific functions in pragma handler to parse for loop, for
example.

Mark's approach isn't clear to me. Is this smth like getting a return
value (some struct) from handlers and attaching it to the next
statement?

-- 
  Dmitry