Subject: Re: A policy for /var
To: Rene Hexel <>
From: Frederick Bruckman <>
List: tech-pkg
Date: 07/27/2002 19:42:54
On 28 Jul 2002, Rene Hexel wrote:

> On Sat, 2002-07-27 at 23:13, Frederick Bruckman wrote:
> > more do-able, would be to move the creation from the install target
> > into the INSTALL script, so that at least a package build machine
> > wouldn't suffer from /var-pollution.
>   That sounds reasonable to me.  BTW, that's what the rpm package
> (targeted by pkg/13058) already does.
> > That drawback is, folks who build
> > from source to run on the same machine would have to remember to run
> > the INSTALL script (but see the "finish" target in www/wwwoffle).
>   What's wrong with a post-install target that runs INSTALL?  I don't
> think packages installed with 'make install' should behave any
> differently than binary packages added with 'pkg_add' (which
> automatically runs the INSTALL script).

Right, but conceptually, it would be a post-package target. Besides,
there's already a post-install target.

>   If we want to make an exception for (bulk) package building machines,
> the way to go IMHO is to put the '${SH} INSTALL' install part between
> '.ifndef(BATCH)' statements (or maybe add some DONT_CLOBBER_MY_MACHINE
> switch if that's what people want).

That's not the only use for it (the target). What I use the "finish"
target in www/wwwoffle for, is to first make the portable package,
then to fix-up the installed files. It could be ".ORDER"'d after
"package", but not depend on "package", so "make install post-package"
would do the right thing.