Re: ADA frontend?

[Samuel Tardieu <sam at ada dot eu dot org>]

>>>>> "Alex" == Alex Buell <alex.buell@tahallah.demon.co.uk> writes:

Alex> I see. Maybe it's time someone started looking into building a
Alex> native ADA front end for egcs.

I guess you don't even imagine how much work it is to write an
efficient Ada front-end. Also, you need to write the standard
libraries, including tasking, which is not going to be easy (Ada has
very well-defined semantics).

You should read GNAT sources (they are very easy to understand) to
have an idea of how much work is required. Ada has been designed to be 
helpful to the programmer, not to the compiler writers :)
 
Alex> That would work best short term, but in the long term we do
Alex> definitely need to look into elegant solutions.

Patch does not mean unstable. GNAT (the Ada front-end) uses GCC 2.8.1
plus some patches for the backend, that will be included in 2.8.2 and
are already in GCC master sources.

Alex> You need ADA to build ADA? Oh dear. It'd be nice if the ADA
Alex> front end was rewritten in C to drop into the egcs system (after
Alex> all that would be quite useful for porting to platforms that
Alex> doesn't have an ADA compiler handy).

As I wrote in response to Jeff, cross-compiling GNAT is very easy,
I've done this many times for embedded targets, and I've also a GNAT
for SparcLinux running with full tasking and distribution support (you 
can program distributed systems in Ada without interfacing yourself
with any socket library or such).

Moreover, GNAT will be available for Alpha/OpenVMS (and is already for 
Alpha/Digital Unix), so if this is your target, you're done :)

  Sam
-- 
Samuel Tardieu -- sam@ada.eu.org