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

Re: tar file, boot floppies



On Thu, 1 Feb 2001, Marcus Brinkmann wrote:

> On Thu, Feb 01, 2001 at 06:53:01AM +0000, Philip Charles wrote:
> > My objective was to get a functioning HURD installation CD(s) out the door
> > and available as quickly as possible. 
> 
> If you can do that without investing too much time, this is helpful
> as an interim solution.
> Otherwise I would suggest to concentrate on the new installer.

The HURD E1 insatllation CDs are on their way to Budapest today.  At the
moment I am working with boot-floppies 2.2.17 as I know they work.

> > Boot Floppies. (I am talking about the Debian package)
> > 
> > Creating base*.tgz (the tarball) is easily separated out from boot
> > floppies and I am treating this separately.  At the moment I am doing some
> > work on this part, and at this stage it seems that it needs to be created
> > on a HURD system.
> 
> Please be more specific about why it seems so. I can only assume that the
> script producing the base.tgz runs some script (probably in a chroot) that
> expect to be able to execute programs from the base files themselve. In this
> case, yes, this needs to run natively.
> 
> One of the things that would be helpful here, btw, is to modify
> tar to be able to read and write translator settings, so we can put the
> "device files" in the base file and have tar installing them at extraction
> time. This is as complex as adding symlink support to tar (which is already
> done), but you don't read/write symlinks but passive translators.

The problem I am encountering is that I get segmentaion faults when I try
to install HURD packages onto the false root partition when using Linux,
despite using --force-architecture.  I was attempting to find a way of
getting the translators into the tarball.  I have not tried --unpack under
Linux.
 
> > Boot/root disks are easy to create for HURD.  They use Linux at the moment
> > and I see no real reason to change to using the HURD.
> 
> There are a couple of reasons, at least: Install and use translators in
> the install program. Proof that the Hurd really is self contained. As the
> root disk is also the rescue disk, having a real Hurd system is not
> unimportant (in the long run).

As the translators will be in the tarball using a Linux ramdisk to unpack
it onto the HURD partition should work.  I agree totally about the HURD
becoming self-contained (the purity issue) and I see the creation of the
tarball within the HURD as the first stage.  To be realy useful the
installation scheme needs to have cfdisk available.

> I definitely want Hurd boot/root disks, but using Linux disks for now is ok
> (the new installer should have hurd disks though).
> 
> I think you are talking about the potato/woody installer, not the new one,
> right? It's okay to butcher that to your liking, as with the new installer
> we will have a better chance to do it right.

Because of my limited programming skills I need to work from a stable
base.  I introduce enough of my own bugs without having to worry about
other people's.  When the new installer comes out in a stable form ...

> > debian-cd (Again I am talking about the Debian package)
> > 
> > The butchered scripts are messy at the moment, and will continue to be so
> > for some time I suspect.  The HURD is very much under development and so
> > the scripts are full of work-a-rounds and these change every few weeks. 

> Yes. There are two issues. Debian stuff needs to seperate between hurd and
> linux all/any packages (the old issue), and the Hurd in general needs to be
> more complete and uptodate.

These are two issues I am not going to have anything to do with, I will
leave them to my betters.

Phil.

-
Philip Charles; 39a Paterson St., Dunedin, New Zealand; +64 3 4882818
Mobile 025 267 9420.  I sell GNU/Linux CDs.   See http://www.copyleft.co.nz




Reply to: