Subject: Re: build.sh defaults file
To: Perry E. Metzger <email@example.com>
From: Curt Sampson <firstname.lastname@example.org>
Date: 02/07/2003 14:48:45
On Fri, 6 Feb 2003, Nathan J. Williams wrote:
> I realize that there is prior art, in the form of /etc/mk.conf, but I
> consider /etc/mk.conf to be 100% a bug, not a feature.
Let me just add my vote that I really would *not* want to see stuff like
this in /etc. Unlike a machine's IP address, how a particular user (of
potentially many on a machine) happens to build one particular source
tree (of potentially many on a machine) is a concern completely separate
from the configuration of the machine on whic the build happens to be
running. I think we already have too many variables to produce reliable
builds anyway (though build.sh is helping to control that now).
(And yes, I've been bitten so many times by my own /etc/mk.conf items.)
My preferred method for this sort of thing is to specify a configuration
file to the build script, e.g., "./build.sh myconfig" or "./build.sh
../myconfig", and to have it use *only* parameter specified by that config
file. This ensures that you know exactly what you're getting when you build.
I also prefer to have the build script fail if not given a config
filename to reduce the chance of accidently building with the wrong
configuration (build with "./build.sh /dev/null" to do a completely
default build), but that's a bit more controversial, I think.
Curt Sampson <email@example.com> +81 90 7737 2974 http://www.netbsd.org
Don't you know, in this new Dark Age, we're all light. --XTC