Re: 3.2 PATCH: Fully support parallel gnat1/gnatbind builds

[Daniel Jacobowitz <drow at mvista dot com>]

On Sat, May 25, 2002 at 12:23:57PM -0400, Robert Dewar wrote:
> > It's interesting that C will bootstrap on pretty much any junk C compiler
> > someone might want to use, but Ada requires one specific version of
> > one compiler, and that's reasonable.
> 
> Not so surprising, C is a very simple language (imagine if all of g++ was
> implemented in C++ :-)
> 
> Also, the sources are a bit junked up to deliberately achieve this goal.
> For example, to me the PROTO calls are annoying noise in the sources, and
> there are a number of places where the use of nested functions would
> help readability and maintainability, but we accept the noise and
> restrictions precisely to achieve the "bootstrap on pretty much any
> junk C compiler".
> 
> In the case of Ada, we early on decided that it was pointless to try
> to maintain compatibility with other proprietary Ada compilers, and
> indeed I have not seen anyone here impeded by that decision.
> 
> And after all, if we had required that GNAT be compilable with any
> earlier version of GNAT, it would be in a very small subset of
> Ada 95 that would be very restrictive, and hamper maintainability.
> 
> One interesting approach would be to have a mode in which GNAT generates
> ANSI C (or actually GNU C would be more reasonable). Then the C sources
> could be used to do a bootstrap. It would be otherwise valuable to have
> a C generating backend for Ada (to bring Ada to machines with a C but
> no Ada), so that would be an interesting project.

Someone from Sun tried to contribute an ANSI C backend (I think, maybe
GNU C...).  As with the last few times this happened, it was shelved by
the FSF; there was some talk of the steering committee discussing it,
but if so nothing public ever came of their conversations.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer