[klibc] Re: Fw: [Bugme-new] [Bug 5627] New: Network boot - IP-Config reports wrong DHCP server address

[Chris Wedgwood]

On Thu, Jan 19, 2006 at 08:38:47AM -0800, Bryan O'Sullivan wrote:

> Yes.  The code in the klibc tree has been sort of ready for that for
> oh, about two years now.

Thus far I'm very happy with it.  I never originally intended to use
klibc but I couldn't get the in-kernel DHCP code to work reliably in
some cases and hacking it further didn't seem attractive at the time
so I tried klibc.

My overblown (contains sutff I don't need/use) initramfs image is a
whopping great 95k --- an indication that klibc is a very powerful
alternative and can also be quite small.  It could be smaller but it
hardly seems worth the effort given that the space is reclaimed via
run-init's transition to the real filesystem.

I suspect much of what I've done here is a duplication of what nash
does.  I'm not really sure.

> What it's lacking is a lot of testing, and it will continue to lack
> it until we make klibc more widely available in testable form
> (perhaps through inclusion in -mm, if Andrew isn't already feeling
> like his edges are bleeding enough).

For me I found the ipconfig client had minor issues.  All dhcp clients
have issues, no too seem to be a like, it's really a horrible mess.  I
could be the dhcp servers here too, I really don't know who it to
blame at this point in time.

It (ipconfig) also doesn't deal with DNS servers (doesn't can't export
them) and klibc right now has no resolver code.

> Did you bring up your modifications on the klibc mailing list?  I
> don't recall seeing anything.

Honestly I haven't yet.  I sorta forgot when I realized the resolver
stuff I did here woulldn't work with the klibc license and I was just
planning on writing a new one from scratch when I got time.

I need to make this more of a priority so I can drop my local
differences.