[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Debian CD mirror on kernel.org



On Mon, 8 Nov 1999, H. Peter Anvin wrote:

> I'm building a Debian CD mirror on kernel.org.  However, I'm unable to
> use the pseudo-image-kit because it requires way too much manual
> intervention.
> 
> What I'd really like to see is something like this:
> 
> 	rsync ... 	# standard Debian mirror
> 	rsync ... 	# Some directory containing the .list files
> 	umask 027	# Don't allow user access until process complete
> 	make-pseudo-images-now-if-anything-changed /pub/mirrors/debian
> /pub/mirrors/debian-cd
> 	rsync ...	# Do the patching to make the pseudo-images offical
> 	chmod -R o+r /pub/mirrors/debian-cd	# Make any new images world
> readable
> 
> What I need is the script "make-pseudo-images-now-if-anything-changed". 
> It seems like an obvious application for "make"; however, the point is
> that I don't want to maintain it.

This can be done quite simple by using the MD5SUMS files, like this:

    cp MD5SUMS MD5SUMS.old
    rsync server::MD5SUMS .
    if diff MD5SUMS MD5SUMS.old ; then
      for each image (or: for each CHANGED image) ; do
        umask/chmod -accessible
        rsync server::image.list .
        make-pseudo-image image.list ftp://server/debian
        mv pseudo-image realname
        rm -rf pseudo-image*
        rsync --block-size=8192 server::realname .
        chmod +accessible
      done
      rm MD5SUMS.old
    fi

NOTE: updating to a "point" release (like 2.1r2 -> 2.1r3) does not necessarily
require the make-pseudo-image; in that case rsyncing the whole tree may be
easier (incl. slink-test which has hardlinks to the actual images (or vice
versa)). However, if you already have a local debian packages mirror, the
pseudo-image way is faster. And of course it's "the only" way to get major
updates (2.1 -> 2.2).

If you have converted this into something useful, please let me know. This is
not really pseudo-image-kit related, but certainly worth publishing on the web
pages.

Regards,
  Anne Bezemer


Reply to: