Re: converting netkit to autoconf?

[Paul Eggert]

> From: Bernd Jendrissek <address@hidden>
> Date: Fri, 13 Sep 2002 17:09:16 +0200
> 
> But what happens when the user configures with CC=uglycc-insane-and-buggy?

Any system-wide cache would have to depend on context like that.
A similar scheme has been used by some vendors' C compilers,
which precompile system headers.  If you compile with nonstandard
flags then you get a different cache.

Another idea for boosting performance, on multiprocessors anyway, is
for Autoconf to generate a makefile that runs the configure-time
tests.  That way, the tests can be run in parallel to some extent.

If we were designing make+autoconf+automake all over again, we'd be
integrating this stuff better, and could get some improvements that
way.  I suspect that adding the proper hooks to 'make' could help us
quite a bit even if we don't fully integrate 'make'.  It would help if
I had time to read up on what the competition is doing (e.g.  Aegis).