A question about dependencies

Mark Linimon linimon at lonesome.com
Mon Jun 28 22:30:17 PDT 2004


On Mon, 28 Jun 2004, Bob Melson wrote:

> To the extent that I believe this to be a problem, it seems to be to
> be one of management and practice.  At the same time, it seems also
> to reflect a degree of, for lack of a better word, sloppiness.

The problem is one of resources to do the necessary regression testing.
i.e. before removing a dependency from a port in the ports tree, one
is obligated to prove that that dependency is not required.  Otherwise
installations of the port can (at least theoretically) break.  The
ports build cluster will be happy to find these and point them out to
you when you do that :-)

In the face of (at the very latest count) 678 ports PRs, and 1837
ports that fail to build in at least one build environment (defined as
cross-product of architecture and release), 170 of which do not even
build on i386-4-stable, and a number of other desired improvements
to the ports infrastructure requiring regression testing (78 PRs for
that alone), the answer is that trimming the dependencies is just
something that generally does not show up on the radar.

This is not to say that these aren't bugs -- they are -- and they
can be checked by sufficiently motivated people by (e.g.) creating a
clean ports jail and installing and running each package with the
trimmed-down dependencies to see if they do indeed still work.  It's
merely a question of where our volunteers want to spend their time
and effort.

mcl



More information about the freebsd-ports mailing list