Re: Source files organisation using gnat

[Geert Bosch <geert@gonzo.sun3.iaf.nl>]

Robert Dewar <dewar@merv.cs.nyu.edu> wrote:
   As to "useless recompilations", actually nothing is uselessly recompiled.
   The internal semantic tree of withed units is reconstructed from source,
   but this is nothing like a full compilation, since the expensive part,
   namely code generation is always skipped for with'ed units.

However many source files are uselessly parsed into a semantic
tree.  When you have a program that uses a large binding, like the
1.2 MB OS/2 API package specifications I use, these will get read
and parsed for every single compilation unit that withs it. The
time spend in parsing these packages specs (8 seconds on my 486DX-100)
tends to be much longer than the actual compilation time. 

The flat nature of these API specifications do lead to big packages,
since splitting them in a hierarchical structure is not meaningful
as you argued a while ago.

So what might be useful is to have some kind of compilation server
that can keep state between compilations and re-use internal datastructures
like parse trees. Of course in that case you may have to worry about
memory management and for example restart the server periodically.

Regards,
   Geert