On 9/21/05, David Woodhouse dwmw2@infradead.org wrote:
On Wed, 2005-09-21 at 14:46 +0100, David Woodhouse wrote:
I think the latter is the bug he wants to report. A workaround is probably to write the contents of the install tree to a DVD and install from that, but it would be better if we could just fix anaconda to install from expanded trees on the hard drive, like it used to.
Oh, the workaround for anaconda's regression, btw, is to _start_ an NFS install from elsewhere, but then switch to tty2 as soon as it's started the second stage of the installer, mount the local file system with the install tree and bind-mount it over the top of /mnt/source.
It's not simple or pretty, and you do still have to have network access to a machine with _parts_ of the install tree -- but it does let you fetch all the _packages_ from your local copy of the tree, instead of over the network. The Fedora/RPMS/ directory can actually be empty on the network server you're using.
gee .. this looks like the dance of the woo-lee masters to me but I will have a go at it
:-)
Actually, it might work if you write a CD containing all but Fedora/RPMS from the install tree, then do the same bind-mount trick to put the RPMS where the installer expects them in /mnt/source/Fedora/RPMS/. At least blank CDs are cheaper than blank DVDs.
I have lots of both. As well as -RW versions which are slow but very handy for testing.
Dennis