Subject: Re: CVS commit: pkgsrc
To: None <tech-pkg@netbsd.org>
From: Joerg Sonnenberger <joerg@britannica.bec.de>
List: pkgsrc-changes
Date: 02/13/2007 22:45:40
Moving it from pkgsrc-changes to tech-pkg.

On Tue, Feb 13, 2007 at 09:53:00PM +0100, Thomas Klausner wrote:
> On Wed, Jan 31, 2007 at 01:26:39AM +0100, Joerg Sonnenberger wrote:
> > >       pkgsrc/devel/bmake: Makefile
> > 
> > Part of bootstrap. Should be marked explictly broken for non-bootstrap
> > usage. On the same note, we should remove the special status of tnftp
> > when it is not build as part of bootstrap.
> 
> Hm, but why shouldn't it build outside of bootstrap too?

Because of the magic for share/mk and the general order issue. This
applies to bmake only, the others can generally be built outside the
tree but might not be installable during bulk builds.

> > >       pkgsrc/graphics/gri: Makefile
> > 
> > ghostscript issue, I was hoping this is fixed.
> 
> Hm, is it?
> Pull the fix up to 2006Q4 and the tag can be removed :)

It worked fine in my last build in head. Not sure what changed to
request a pullup.

> > >       pkgsrc/security/botan: Makefile
> > 
> > Feature test, not sure how to best handle this.
> 
> Disable it?

I don't think it is a good idea to disable TSC support, which is IIRC
the culprit. Or something like that.

> > I object those, as they build fine at least on DragonFly and most also
> > with modular Xorg.
> 
> The test case currently is the latest NetBSD release, i.e. NetBSD-3.1/i386.
> I'm aware that some of them build on NetBSD-current, too.
> 
> They are not marked as BROKEN, so I'd let the tags stay.
> 
> Or we should reconsider what our intention with the tags is...

For the tag should form the basic to remove broken packages by threating
people accordingly. Broken packages is somewhat wishy-washy, some rules
should be defined. A good suggestion for me includes:
- breaks in bulk builds of both DragonFly and two NetBSD releases
- has no special reason for the breakage (see above for examples :-))
- breaks for obvious stupidness (hi p5-math-pari!)
Packages which fail the first point are often a good candidate for
simple fixes or platform masks, but not always.

Joerg