Re: Proposed changes to _bsd_pkg -- request for comments

[Roman Neuhauser <neuhauser@bellavista.cz>]

# j-devenish@users.sourceforge.net / 2003-01-03 16:39:27 +0800:
> I am looking for some feedback from *BSD users/admins who find
> themselves using zsh's _bsd_pkg completion for pkg_add, pkg_create,
> pkg_delete and pkg_info. I am working with OpenBSD but would really
> appreciate some feedback from someone using _bsd_pkg under FreeBSD.

    I don't really use pkg_*, but do use FreeBSD. more below.
 
> Includes a patch for _bsd_pkg_pkgfiles(), which was matching paths
> against an undefined variable $PKG_PATH. Is this defined under FreeBSD,
> perhaps? I found the completion worked erroneously under OpenBSD.

    pkg_add(1):

    The value of the PKG_PATH is used if a given package can't be found.  The
    environment variable should be a series of entries separated by colons.
    Each entry consists of a directory name.  The current directory may be
    indicated implicitly by an empty directory name, or explicitly by a sin­
    gle period.

    I don't use the pkg_* tools, but looks like it should be
    ${PKG_PATH:-/usr/ports/packages} or something like that.

> My attached version fixes that, and also allows this sort of completion:
 
> % pkg_add <Tab>
> file1.tgz foo.tgz bar.tgz
> % pkg_add dir1/<Tab>file3.tgz
> % pkg_add databases/<Tab>
> db1.tgz db2.tgz
> % pkg_add foo<Tab>
> pkg_add /usr/ports/packages/blah/All/foo.tgz
> % pkg_add -v foo.tgz
> pkg_add -v /usr/ports/packages/blah/All/foo.tgz

    nice!

> I like the "new" behaviour.

    so do I. :)

> Basically, the selection list is much more readable and can understand
> packages subdirectories (to help jog one's memory). So far, the latter
> is in a 'case' construct for OpenBSD since I don't know how other BSDs
> lay out their /usr/ports/packages/*.

    should be the same.
 
> Under OpenBSD, the -a option can only be used on its own (no other flags
> or arguments are appropriate). I have implemented this inside a 'case'
> for openbsd*. If the other BSDs are the same, then we can reduce some
> of the 'case' constructs.

    A quick test shows that -v, -p, -q, -g further modify the -a output,
    others are noop. I haven't found a switch that would make it break
    (tried about 1/2 of them)
 
-- 
If you cc me or remove the list(s) completely I'll most likely ignore
your message.    see http://www.eyrie.org./~eagle/faqs/questions.html