[Pkg-uml-devel] [VERY LONG - uml planning for etch] how to plan the future

[Andreas Schuldei]

On Thu, Jan 12, 2006 at 11:54:27PM +0100, Stefano Melchior wrote:
> Hi all,
> premitting that is a lot to do, let't start to resume some considerations.
> The pkgs to take care are:
> 
> 1 - user-mode-linux;

i use this

> 2 - user-mode-linux-doc;
> 3 - uml-utilities;

and this

> 4 - rootstrap.

and this alot

> 5 - kernel-patch-uml;
> 6 - kernel-patch-skas;

and this NOT, since i always pull stuff directly from
Blaisorblade's page.

(just to let you know where i am standing)

> 7 - umlrun;
> 
> Let's take in consideration one at the time:
> 
> (1) user-mode-linux
> It is crucial. At the moment there is the support for 2.4.27, 2.6.8 (but
> not working) and 2.6.12. This could be enough, even though its support is
> in deplay, for Sarge. Etch requires at least 2.6.14 support.
> What about the `old` 2.4.27? what 2.4.x serie kernle will be supported on
> Etch?

the goal is to have no 2.4 in etch, aparently. which is very good
for us.

> So it is time to start from the 2.6.12-1um idea and implement the
> 2.6.{14,15} support. That is the target. 2.6.12 is out of date, now.
> I subscribed the uml lists on sourceforge and 2.6.15 is not so stable and
> reliable to be definitively included. So first, my idea, is to focus on
> 2.6.14, waiting for a better release.

what about trying to cooperate with the kernel team and supply
them with fixes to make the kernel that seems to end up in etch
to make uml run as good as possible? will such patches be
produced by the uml hackers for e.g. 2.6.15?

> (3) uml-utilities
> At the moment is in sync with the official u-m-l support.
> http://prdownloads.sourceforge.net/user-mode-linux/uml_utilities_20040406.tar.bz2
> It needs some bugs (a couple tagged as serious: #326287, #326029) to be fixed.
> On svn you can find a new release with one of these fixed.
> 
> (4) rootstrap
> I have an idea of how to fix the problem of modules copy on root_fs, but I
> first would like to test it. 

which problem is that? i am not aware of any problems with that.

> I skip it for the weekend.
> Since the author is Matt, while the other distros adopted their own way of
> creating a new root_fs, it could be a good idea to see if we can take
> advantage of some other tools good ideas to improve rootstrap. Do you
> think it is possible?

i am not sure rootstrap is so bad in comparison to what other
distros have. i saw some static scripts for rh and rootstrap's
modules give you a lot more control and flexibility then that.
(plus that i invested  alot of time into it already and would
weep if it was flushed down the drain. (c:)


> (5) kernel-patch-uml
> It is time to choose the future of this pkg: it should not be usefull
> since the kernel version newer that 2.6.8 provides um architecture support
> embedded. But we still need to support the previous kernel ones, so we are
> encouraged to keep it alive and, above all, to fix the bugs.

really? why? for etch (what we work towards) it is not really
desireable to have a wide range of all sorts of kernels
supported. people pick debian because they just installed a
pre-compiled package and it works. they dont want to fiddle much
with it. 

if we manage to work with the kernel guys and make their kernel
compile a stable, good uml kernel and we can ship that
precompiled we are all set, i think. 

should we try to get the skas kernel into the mainline kernel?
does it have any penalites for the non-uml case?

> (6) kernel-patch-skas
> It is not the same of the previous package, because it is necessary for the host kernel.
> But I heard about issues for 2.6.{14,15} kernel version; at the moment it
> seems not so stable or reliable.

i am running two boxes with 2.6.14 with skas and i see no issues
with it. what is the problem?


could you please give the URLs of the svn and the cvs archive?
i mean the output of svn info and cvs's CVS/Root content.