FreeBSD 9.1-RC1 Available...

Peter Wemm peter at
Thu Aug 23 22:14:28 UTC 2012

On Thu, Aug 23, 2012 at 2:52 PM, Simon L. B. Nielsen <simon at> wrote:
> On 23 Aug 2012, at 20:41, Peter Wemm <peter at> wrote:
>>>> On Thu, 23 Aug 2012 00:50:46 -0400, Ken Smith wrote:
>>>>> With both the doc and ports repositories now moved to SVN it has been
>>>>> decided to not export the 9.1 release branch activity to CVS.  So
>>>>> csup/cvsup update mechanisms are not available for updating to 9.1-RC1.
>>>>> If you would like to use SVN the branch to use is releng/9.1.
>> * RIght now you can mirror
>> svn://{stable/9,release/9.1*,etc} to your laptop,
>> and Simon is setting up a US-east coast and US-west coast mirror.  You
>> can easily switch your mirrors on the fly if there's a closer/faster
>> one.
> Real world got in the way, but they should be ready by Friday or Saturday, and will be documented in the Handbook.
> On a side note, the svn part of one mirror had to be re-created a few days ago as svnsync gets very unhappy if you pull the power while it's running (we had a PSU die). There seem to be a fair chance your local repository gets corrupted enough in that case that svn just gives up. Not that it's a big problem, but is a bit annoying.
> We (clusteradm) are working on getting an EU mirror up, but no timeframe yet.
>> You can do your own personal svnsync all the way from rev 0, but it
>> takes some time.  It's more time efficient to start with a seed and
>> let it catch up.
> And if people are wondering, some time is really some time. The original ports mirror took ~24h to create - and that was over the LAN. Don't do that yourself unless you feel like doing a latency test of your internet connection :-).

You'll also find out very quickly how much fsync(2) hurts on a softdep
or su+j system.  The svn fsfs backend does a fsync multiple times per
revision to guarantee its transaction boundaries.

There's a reason why I wrote nofsync.ko for these scenarios.  Back
when I did the first mirror for making a seed, it was the difference
between what was shaping up to take about 12 hours vs what ended up
taking 25 minutes with nofsync.ko + noatime + full async + no softdep
+ no softdep+j.

Peter Wemm - peter at; peter at; peter at; KI6FJV
"All of this is for nothing if we don't go to the stars" - JMS/B5
"If Java had true garbage collection, most programs would delete
themselves upon execution." -- Robert Sewell

More information about the freebsd-stable mailing list