preferred place for system-wide config files

Ion-Mihai Tetcu itetcu at
Tue May 18 15:53:47 UTC 2010

Hash: SHA1

On Tue, 18 May 2010 13:13:36 +0100
Matthew Seaman <m.seaman at> wrote:

> Hash: SHA1
> On 18/05/2010 12:56:11, Wesley Shields wrote:
> > On Tue, May 18, 2010 at 07:29:22AM +0400, Anonymous wrote:
> >> a little confusion arised from ports/146627.
> >>
> >> The Porter's Handbook defines PREFIX as the place where port should
> >> install its files including config files. However, it doesn't say
> >> where the port should try to read its config files that may not
> >> exist: LOCALBASE/etc or PREFIX/etc? Until recently I assumed such
> >> files as installed together with the port unless they belong to
> >> some other port.
> >>
> >> For example, many GNU_CONFIGURE ports look for config files under
> >> SYSCONFDIR that's usually under PREFIX/etc and some of them don't
> >> install config files nor samples. A few examples: xorg-server,
> >> subversion, git. Should such ports be modified to use
> >> LOCALBASE/etc?
> > 
> > PREFIX/etc is the correct place. Very few ports touch LOCALBASE at
> > all, and when they do it's explicitly requested. I'm thinking of
> > bind here as an example.
> The principle is that PREFIX is where *this* port is going to be
> installed, 


> but LOCALBASE is where you should assume any prior dependency ports
> have been installed. 


> PREFIX and LOCALBASE are virtually never set to different values
> although I do find doing that useful for tesing ports etc. during
> maintenance work.

Two examples when they differ:
- - when you install a port as user under your home dir
- - whe you need multiple and conflicting versions of the same port

> I know there have been Qat port building experiments setting both
> PREFIX and LOCALBASE to a non-standard value,

Yes. And it's a new instance is in work.

> but I don't think there have been any tests trying to install each
> port to a PREFIX different to the LOCALBASE used for all its
> dependencies.

No, not yet, because of limitation of the current testing

> It's also the case that some ports have to match the PREFIX/LOCALBASE
> of certain dependencies: for example various of the RT extension
> modules I maintain are like that.

Yes, they are basically at fault (except some really really corner

- -- 
IOnut - Un^d^dregistered ;) FreeBSD "user"
  "Intellectual Property" is   nowhere near as valuable   as "Intellect"
FreeBSD committer -> itetcu at, PGP Key ID 057E9F8B493A297B
Version: GnuPG v2.0.14 (FreeBSD)


More information about the freebsd-ports mailing list