[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 01 Dec 2012 01:01:28 -0600
From: Rob Landley <rob@...dley.net>
To: musl@...ts.openwall.com
Cc: sabotage@...ts.openwall.com
Subject: Re: [musl] [ANN] sabotage mips(el) 0.9.7+ rootfs tarballs are
available
On 11/30/2012 08:40:15 PM, John Spencer wrote:
>
> this is based on musl 0.9.7, with all mips patches applied.
> so it really is more or less 0.9.8 minus the ppc port.
>
> it takes about 2-3 weeks to get everything built in qemu mips,
> so i didnt restart the build when 0.9.8 came out.
Wow, really?
Beyond distcc, I looked at cluster based approaches but they all
leverage distro repositories. Basically you have a "master" node that
hands out work to "slave" nodes which start out all using the same
filesystem, the master works otu which packages it can build with out
dependencies and hands them out to the "slave" nodes, which build them,
create a binary package (.rpm, .deb, the portage binary thing, etc) and
sends it back to the "master" node which loads them into the
repository. The master then hands out more work to the idle slave nodes
using only packages which have prerequisites that can be installed from
the repository rather than built.
(I looked into having them all share a network root filesystem and
having the master node update it, but running ./configure while a
package shows up out from under you is kinky.)
I also covered setting up tun/tap for kvm/qemu here:
http://landley.net/lxc/02-networking.html
(That covers more than you'd need, but if you have a 4-way machine you
can get a small cluster setup that way, and each qemu instance gets a
cpu to itself...)
Rob
Powered by blists - more mailing lists
Powered by Openwall GNU/*/Linux -
Powered by OpenVZ