Subject: Re: make depend. a good idea, but is it the law?
To: Chris G. Demetriou <firstname.lastname@example.org>
From: Todd Whitesel <email@example.com>
Date: 01/23/2000 15:52:10
> This is bogus. if you _ever_ need a 'make depend' before a build of a
> clean source tree, somebody has broken the source tree. (Further, if
> there's no way to do a complete build without also building
> dependencies, something is broken as well.)
Ah, refreshing news.
You've just doubled the rate at which I can build kernels on slow boxes.
With the poor-man's cross build setup I have going, ``cc -c'' time is
beginning to approach ``cc -E'' time for the same file.
It is now preprocessor limited, and ``make depend'' spends all of its
time in the preprocessor building that list of included files, only to
throw out the .i output because it wasn't requested.
What I would love to see is an option to the preprocessor that acts like
``cc -E'' but also deposits the mkdep information into a separate file,
or (better) writes out the .i file and dumps the mkdep info to stdout.
Then the ``make depend'' step could leave .i files all over the place,
and we could have a variant of ``make all'' that goes explicitly through
the .i files, so that if they are still present from a depend, we don't
bother to preprocess them again.
Of course, ``make dependall'' can use it too.
toddpw @ best.com