Re: parallel translation

[Georg Bauhaus <rm.tsoh+bauhaus@maps.futureapps.de>]

On Tue, 2007-11-20 at 18:47 +0000, Jeffrey R. Carter wrote:
> Georg Bauhaus wrote:
> > 
> > P1 (in file "p1.ads", say) depends on P2 (in file "p2.ads", say).
> > I am assuming that Task 1 is supposed to compile P1, and task 2
> > is supposed to compile P2. An elaboration order will have to be
> > established in order to translate almost anything. Here, for
> > example, task 1 (compiling P1) must know what is in P2 (compiled
> > in task 2) because P1 depends on P2. How do the compilers etc.
> > communicate?
> 
> No, GNAT is source based. The compilation of P1 will read p2.ads; the 
> compilation of P2 will read p2.ads and p2.adb (if needed). The compilation of P1 
> need not wait for P2 to be compiled.

Yes, I had felt uneasy after posting as the comment confused translation
and elaboration. Yet, I had understood that something like a shell
script was to control starts of gccs for each file, basically. What is
the win over "gnatmake -jnnn" when none of the independent compilers
knows which file has already been processed? There will not normally
be a "p2.ali" after "gcc -c" has read and analyzed "p2.ads" (with'ed
by P1).

Time to look at the sources of gnatmake.