On Wednesday 04 July 2001  2:35 am, Greg A. Woods wrote:
> [ On Tuesday, July 3, 2001 at 23:05:14 (+0100), Chris Gilbert wrote: ]
>
> > Subject: Re: build of -current from today fails on libexec/kpasswdd
> >
> > On Tuesday 03 July 2001 12:06 pm, Chris Pinnock wrote:
> > DESTDIR must be in the enviroment, there's a script somewhere that needs
> > it in the enviroment.  Maybe it only effects make release, as I can't
> > remember the exact script.
>
> No, DESTDIR does not have to be in the environment.  As far as I can
> tell the only scripts that require it in their environment have their
> environments explicitly crafted by the make rules.  For example in
> distrib/sets/Makefile:
>
> 	checkflist: check_destdir
> 	        DESTDIR=${DESTDIR} sh ${.CURDIR}/checkflist
>
> Any that don't do that are broken and need fixing (though I don't think
> there are any as I've been doing builds for the past two weeks with
> narry a problem, at least not in that department :-).

kdump and makewhatis appear to need DESTDIR in the env, some arch's 
floppy/ram disk image creation scripts need it as well.

Things in ./gnu/dist/toolchain/gcc/fixinc seem to have a DESTDIR (but I don't 
think it's the same as the BSD one, and we don't currently use that version 
of toolchain)

> Not only that, but it's best not to set it in the environment lest you
> forget it is there and it gets used at the wrong time.  The environment
> is a necessary evil best to be avoided where possible, especially in
> software configuration management!

I'm not saying it's the right thing, I'm saying that currently we've got to 
have it in the env to make sure things are done right.

I got bitten by this when building 1.5.x while running -current, and also 
when building -current when running 1.5.x as kdump gets confused :)

Cheers,
Chris