Well here it is, a giant pile of patches that completely reworks all the
installation repo selection and stage2 finding. Some of these patches are
probably a little too small, and some are probably a little too large. But
this is a huge body of work, so that's just how it goes.

Current status:

- The entire purpose of method selection in loader now is trying to find the
stage2 image. If the stage2 image can be found on an attached CD, that's
what will get used.
- You can specify where the installation repo is located with repo= or method=.
The intention is that method= is deprecated and will go away in the future,
but we need to preserve it at least for now.
- stage2= still exists and will continue to exist. preupgrade probably doesn't
work, though. If you give a stage2=, that takes priority over any image
that might be found on the CD.
- If you don't give a method= or repo=, we first look for any CD/DVDs. If
one is found, we look for packages. If those are found, we use them. If
not, we default to using the mirrorlist provided by the /etc/yum.repos.d
files. This means that for the most common installation method, you'll
just toss your boot.iso in the drive, run straight into stage2, and be up
using the mirrorlist. Hooray!
- Since very few people will be getting the method selection screen in loader,
most people won't have networking up when we get to stage2. That means in
order to use the mirrorlist, we have to use the NetworkConfigurator. This
isn't quite ideal, but David's going to have a giant networking patch one
of these days so I'm not really going to bother.


Current problems:

- Media check is commented out. My intention here is to move this to stage2
if possible.
- yum logs to tty1.
- HDISO and NFSISO are completely untested and probably don't work, but I'll
make them work later.
- We don't write out a method to /root/anaconda-ks.cfg.
- Error handling in AnacondaYum.__init__ is rather poor, but I need a better
repo editor to fix that, and that's on a whole different branch I need to
return to.


Anyway this isn't 100% working but it seems good enough for me to consider
committing. I've tested boot.iso and pxebooting with HTTP, NFS, and a DVD
and those all worked decently. The big problem for me right now is the error
handling. But that can be beefed up once this stuff goes in and I have an
easier time testing.

Questions?

- Chris

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

On Tue, 2008-06-17 at 11:45 -0400, Chris Lumens wrote:
> - Since very few people will be getting the method selection screen in loader,
> most people won't have networking up when we get to stage2. That means in
> order to use the mirrorlist, we have to use the NetworkConfigurator. This
> isn't quite ideal, but David's going to have a giant networking patch one
> of these days so I'm not really going to bother.

I don't know that this is worse than the current state, though. The
NetworkConfigurator has the advantage of being accessible to people in
more languages

> Current problems:
> - Media check is commented out. My intention here is to move this to stage2
> if possible.

So the question I'd have about this is what do we do to ensure that
reading the stage2 isn't bad. But maybe that's less of a problem these
days.

> Anyway this isn't 100% working but it seems good enough for me to consider
> committing. I've tested boot.iso and pxebooting with HTTP, NFS, and a DVD
> and those all worked decently. The big problem for me right now is the error
> handling. But that can be beefed up once this stuff goes in and I have an
> easier time testing.

It sounds good to me from the high-level. And the patches look pretty
reasonable from a quick look through them

Jeremy

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list