Re: IPFILTER Branch [was: Re: [current tinderbox] failure on powerpc/powerpc]

Darren Reed wrote:
> On Sat, May 07, 2005 at 09:08:52PM -0600, Scott Long wrote:
> 
>>Darren's latest import wasn't ever tested on FreeBSD, nor did he make
>>any attempt to make it compile on FreeBSD before he imported it.
> 
> 
> That's just utter bullshit.
> 
> 
>>He merely dropped it in CVS and walked away.
> 
> 
> As is that.  I spent the best partof the next 2 to 3 days getting
> it together.  But I could be generous and accept that it appeared
> like that but really, what you call "walking away" is what I call
> "sleeping."
> 
> Maybe you have more time in your life such that you can do commits
> in hours that aren't next to when you sleep.  I'm not so fortunate.
> 
> Or maybe you just don't need to sleep.
> 
> Or maybe you're just ignorant that people do work when they can and
> in timezones that don't align well with most European/North Americans.
> 
> 
>>If it wasn't for the teamwork of others, it would still be broken.
> 
> 
> This time there were some parts that needed work that were outside my
> experience for FreeBSD.  That's unlikely to be needed again now that the
> framework has changed.
> 
> 
>>CVS is where work goes that is done and tested, not where undercooked
>>hacks go while waiting for others to clean them up.
> 
> 
> It's always tested (as in it works) before it is imported.
> 
> The only issue I generally have is with driving CVS and making sure all
> the right bits go in the right place and that it compiles cleanly.
> 
> If I had enough "stuff" I'd rsync out the repo, do a test of everything
> against a clone of the repo, test that a few times and then commit into
> FreeBSD.  Well, at least for the last import.  But the only resource *I*
> have (currently) are those in freebsd.org and my laptop that needs to be
> used for a lot more things (like run FreeBSD 4 and  FreeBSD 5, too.)
> 
> Darren

I completely understand that integrating IPFilter into FreeBSD is not an
easy process.  I also completely understand that it's impossible to do a
vendor import with CVS and have the result work immediately.  Vendor 
imports are often a process of import-then-fixup.  That's fine. 
However, the fact that it took 3 days to do the fixup portion points to
the fact that no integration work was done before hand.  Before you hit
the commit button, you should have already known how to fix the hard
problems that you encountered.  You should have already put everything
in place on either  local or remote machine and gotten through a
buildworld.  But you didn't, you assumed that you could do the import
blindly and that there would be few or no problems.  That cost us 3
days of the tree being broken.  And that is completely unacceptable.
Sun wouldn't accept their tree being broken for three days.  The tools
and the resources exist within FreeBSD to avoid this, there is simply
no excuse to flagrantly disregard the minimum standards that we have.
CVS is not a sandbox, not even the HEAD branch.  If you want a sandbox
then use Perforce, or at least use a projects/ branch in CVS.  That's
about all I'm going to say about that.

Scott