1 |
On Thu, Jun 14, 2012 at 01:49:16AM +0530, Yohan Pereira wrote: |
2 |
> On Wed, Jun 13, 2012 at 08:43:20AM +0200, Daniel Wagener wrote: |
3 |
> > On Wed, 13 Jun 2012 01:04:19 -0500 |
4 |
> > |
5 |
> > What about a different approach: gentoo in a VM on the desktop |
6 |
> > Would that not be much easier? |
7 |
> > Of course some processor power is used for the VM itself, but it should still significantly decrease compilation time on the laptop. |
8 |
> > Plus you can easily equip other machines with that VM and use their power too. |
9 |
> > |
10 |
> What about a chroot? that would be much lighter compared to a VM. |
11 |
> Alternativly you can use the chroot to build binnary pkgs that can be |
12 |
> installed on the laptop. |
13 |
|
14 |
There are 3 solutions I can think of: |
15 |
|
16 |
1. Get a cross compiler for i686 working under fedora (have no idea |
17 |
about that, except manually setuping and compiling gcc) and setup the |
18 |
Fedora's distcc to use that correctly (this should go more or less |
19 |
according to the docs on the gentoo wiki regarding cmake and i686 vs |
20 |
amd64) |
21 |
|
22 |
2. Create a gentoo chroot on the Fedora OS. Much "lightweight" that a |
23 |
virtual machine, you can setup the cross toolchain and distcc according |
24 |
to the wiki, you just have to play a bit with how to start the distcc |
25 |
inside the chroot (a plain /etc/init.d/distcc start inside the chroot |
26 |
won't work). Note that you can run a i686 chroot in an amd64 system, so |
27 |
you actually don't need to set up any fancy crosscompiler inside the |
28 |
chroot. Just run the chroot as 'linux32 chroot /mnt/chrooot ...' to make |
29 |
sure uname and similar get correct info... |
30 |
|
31 |
3. Mount the laptops root filesystem through nfs on the fast computer |
32 |
(use no_root_squash on the laptop export to have correct root access to |
33 |
files) bind-mount something local (disk or tmpfs if you have enough mem) |
34 |
over /var/tmp/portage, chroot into it (don't forget to mount /proc, |
35 |
maybe /sys and maybe bind-mount /dev, though that should not be needed |
36 |
and don't forget 'linux32 chroot') and run emerges there... you will |
37 |
actually be running everything on the fast computer, only the access to |
38 |
the laptops disk will be through the network. With a fast network it |
39 |
should be a lot faster then working ont the slower notebook (note that |
40 |
if you bind-mount /var/tmp/portage inside the chroot, most of the |
41 |
compilation will be working with a local disk...) Compared to the distcc |
42 |
method, even the configure phases will be much faster... I do this |
43 |
often with my Pentium M 1.6Ghz thinkpad laptop and my quad core amd64 |
44 |
desktop... |
45 |
|
46 |
|
47 |
|
48 |
yoyo |
49 |
|
50 |
> |