On Sun, 27 Aug 2006, M. Warner Losh wrote:
> In message: <20060826082751.GA11942@freebie.xs4all.nl>
> Wilko Bulte <firstname.lastname@example.org> writes:
> : On Fri, Aug 25, 2006 at 07:31:20PM -0700, Steve Kargl wrote..
> : > On Sat, Aug 26, 2006 at 12:08:54PM +1000, Peter Jeremy wrote:
> : > > On Thu, 2006-Aug-24 18:44:02 -0400, Alexander Kabaev wrote:
> : > > >personally. The mistake you made just shows how important it is to do a
> : > > >buildworld _from freshly checked out sources_ after every major vendor
> : > > >import.
> : > >
> : > > I'm not sure how doing this would have prevented the breakage. It
> : > > might have allowed Guido to notice the breakage more quickly.
> : >
> : > It would have caught the first problem, which I reported here
> : > http://lists.freebsd.org/pipermail/freebsd-current/2006-August/065175.html.
> : > This problem was fixed by kan as noted here
> : > http://lists.freebsd.org/pipermail/cvs-src/2006-August/067536.html
> : >
> : > Notice I reported the problem of Aug 16th. Kan's patch was committed
> : > on Aug 17th.
> : >
> : > > Guido made a mistake, has apologised and has corrected the problem in
> : > > less than 6 hours.
> : >
> : > He only corrected the second problem caused by his commit.
> : Can we please stop this moaning about Guido's work? Mistakes do get
> : made, and in this case were fixed rapidly.
> Yes. The number of commits that I've actually pulled a fresh world
> and built after are in the low double digits (compared with ~3000
> total commits). The number of times I've broken the tree is also in
> the low double digits. I don't find the return on investment of
> always doing a full build on a fresh tree worth the effort. I think
> I've done a full make universe after a set of commits twice, and those
> were esepcially risky changes.
> People need to understand that mistakes happen from time to time.
> These mistakes do cost other people time, so they should be avoided as
> much as possible, and forgiven when they happen. We do not have a
> major problem with tree breakage, despite what some would say. The
> tree usually builds, and usually works. It is the few times that it
> doesn't that stand in stark contrast...
That's not entirely true. There are often periods of extended
carelessness. I know that because I was building the tree 3-4 times a
week as a minimum to support RE activities as well testing my own changes.
Guido's mistake here in RELENG_6 is forgivable given the circumstances,
but in general Alexander's frustration over the low quality of commits to
CVS is entirely justified. Your comments are noted, but do nothing to
ensure good quality and engineering practices.