Re: seperate keyword and seperate compilation with Gnat?

[dewar@cs.nyu.edu (Robert Dewar)]

Jon says

"So, these bits seem to indicate that GNAT is playing a little loose with
the rules. But -"

Nope, not true! You are making a fundamental assumption that is wrong,
namely that compilation as defined in the RM corresponds to compilation
with the default options using gcc.

No one told you this -- you just assumed it :-)

If you want to get formal, the compilation process in the RM corresponds
to compiling with the -gnatc switch, then the post-compilation step
can be done with gnatmake. 

In fact we use exactly this approach for some of the complex multi-file
tests in the ACVC suite.

So you can discuss if you like the pragmatic implications of the GNAT
approach, as Bob and I have already done, but don't spend time trying
to see if the RM allows a compiler to require that subunits be present
when the parent unit is compiled, it definitely does NOT, but neither
does GNAT, when you use the proper compilation approach.