Re: Update on fossil repositories

[bch%methodlogic.net@localhost]

On Thu, Mar 31, 2011 at 12:22:16AM +0200, Joerg Sonnenberger wrote:
> On Wed, Mar 30, 2011 at 09:58:09PM +0000, bch%methodlogic.net@localhost wrote:
> > I think the prev pkgsrc version was *not* delta manifest'd though... the
> > repo is ~588M, 60935 files, versus netbsd-src's 3.2G, 127280
> > files... rebuilding pkgsrc too ~6x as long as netbsd-src, iirc.
> 
> The numbers are a bit misleading. src has twice as many artifats and and
> 30% more check-ins than pkgsrc. But -- and that's the critical part --
> src tends to have very short delta chains. The majority of the cases are
> covered by the small delta cache, so e.g. the manifests tend to have the
> base artificat around in the cache already. Especially doc/CHANGES* in
> pkgsrc is quite different though. I suspect that's a large part of the
> longer rebuild time.

What you're talking about above may well be true; I'm not in a position to
comment. Querying the repo databases directly indicates that my initial
suspicion is indeed true, though:

kamloops$ pwd
/home/bch/work/netbsd-src
kamloops$ fossil sqlite3
SQLite version 3.7.6
Enter ".help" for instructions
Enter SQL statements terminated with a ";"
sqlite> select value from config where name='seen-delta-manifest';
1
sqlite> kamloops$ cd ../pkgsrc/
kamloops$ fossil sqlite3
SQLite version 3.7.6
Enter ".help" for instructions
Enter SQL statements terminated with a ";"
sqlite> select value from config where name='seen-delta-manifest';
sqlite> kamloops$

Anyway, can take this offline if you like. Just interested in seeing
the repos representing themselves as best they can.

Cheers,

-bch

-- 
Brad Harder
Method Logic Digital Consulting
http://methodlogic.net/
http://twitter.com/bcharder