committed: Ada updates

[Diego Novillo]

On Thu, 2004-04-08 at 11:47, Richard Kenner wrote:
>     The trick is to make sure that you remove all the semantics from the
>     input language so that the GIMPLE optimizers don't munge things up. 
>     Since I don't know anything about Ada it is very difficult for me to
>     estimate.  I'm just guessing.
> 
> That's certainly what's happening.  Look at the code I pointed you to.
> The tree nodes currently being used are defined in language-independent
> terms.  They are implemented by calling the standard expand_* routines.
> 
Yeah, I was just browsing through that.  OK, so then it should only be a
matter of not calling expand_* and generate the equivalent GENERIC nodes
instead.  Longer term, the Ada parser could generate GENERIC nodes
directly, but native parse trees may have other uses, so I wouldn't
recommend that.


> Note that the majority of Ada semantics have already been removed by the
> part of the front end written in Ada before even calling the C code at all.
>
OK.  That's good.  In principle, it doesn't seem that it will be too
hard.  We'll see.


Diego.