Subject: Re: pkgsrc NetBSD 3.99.18/x86_64 bulk build results 20060501.0807
To: None <pkgsrc-bulk@netbsd.org>
From: None <martijnb@atlas.ipv6.stack.nl>
List: pkgsrc-bulk
Date: 05/22/2006 21:32:10
Thomas Klausner:

> Some other packages wanted to pull in imlib, but this failed too.
> I'm not quite sure why, imlib is not marked as broken in the bulk
> build output.

I've seen this happen in the past too, when there's *massive* outage. When
devel/glib2 was broken, it had the same symptoms: A broken package that 
isn't being recorded as such in the build report, but which will still make
dependent packages to fail.

In this case, imlib failed because netpbm broke:

imlib-1.9.15nb4 (graphics/imlib) is broken because it depends upon
netpbm-10.33nb2 (graphics/netpbm) which is broken.

netpbm broke (pkg/33347, fixed in the mean time), and apperently failed to 
register this properly:

BULK> Marking all packages which depend upon netpbm-10.33nb2 as broken:
<ul>
<li>mgetty-1.1.31nb5 (comms/mgetty+sendfax)</li>
<li>avr-libc-1.2.5nb2 (cross/avr-libc)</li>
<li>aview-1.3.0.1nb7 (graphics/aview)</li>
[...]
<li>gkx86info-0.0.2nb6 (sysutils/gkx86info)</li>
cannot create /bulk/sysutils/gkx86info/.broken.html: directory nonexistent
BULK> Build for netpbm-10.33nb2 was not successful, aborting.
</body>
</html>

Aperenly, sysutils/gkx86info somehow caused it to abort further reporting
and bookkeeping. When I checked, /bulk/sysutils/gkx86info was present, but
it must've gotten created at a later stage.

Martijn.