svn commit: r189594 - head

[Mark Linimon]

On Mon, Mar 09, 2009 at 03:27:05PM -0700, Steve Kargl wrote:
> When Mark Linimon, a member of portmgr, posts on Feb 26th 
> 
> (http://lists.freebsd.org/pipermail/freebsd-ports/2009-February/053282.html
> 
> There appears to be a disconnect with USB2 development and the
> rest of FreeBSD.

Within 2 days after posting that, we had a list created of all the
ports the ports that were broken by known commits to -current, and
people started working through them.  I am very encouraged by the
response.

In theory, there would be enough volunteers and spare machines to
regression-test every change that might be disruptive before commit.
In practice, it's simply not possible.  As it is, since 7.1 went out
the door we (portmgr) have been continually running -exp runs to try
to work through our backlog.  We have several more in the queue that
are also high priority.  (In particular, we spent time on the xorg
upgrade, which we knew would affect all users, not just -current).
I suppose portmgr could have objected to the merge -- but note, it
had been in the planning stage for quite some time, and the window
to get this massive change into the src base before 8.0 was starting
to close.

Now that the commit has gone in, all we can do is ask people to help
in fixing problems.  In this, I think we could hardly have done
better.

As mlaier has pointed out, -current has sharp edges.  It's one of
3 choices open to you, the other two being -stable (which will still
have ports regressions from time to time -- see xorg -- and sometimes
even src regressions), and a release, which is the best we can do with
respect to QA.  If you can't deal with having your system out of
commission on occasion, then -current isn't for you.

mcl