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

Re: [Debootloaders-yaboot] Bug#587290: How does linux-base modify /etc/yaboot.conf ?



Comments in-line below...

On Sep 18, 2010, at 1:34 PM, Ben Hutchings wrote:

On Thu, 2010-09-16 at 15:44 +0200, Benjamin Cama wrote:
Le jeudi 16 septembre 2010 à 15:12 +0200, Benjamin Cama a écrit :
You mentioned that linux-base may modify /etc/yaboot.conf. I think it might be the culprit for this bug, but I would first like to know how
and why it does that ?

OK, I was right, thanks to your hint: see
http://svn.debian.org/wsvn/kernel/dists/sid/linux-2.6/debian/linux-base.postinst

OK.  The final (I sincerely hope!) piece to the puzzle.


The yaboot.conf is taken as if it has the lilo syntax.

It does. Compare the function cfg_get_token() between lilo (cfg.c) and
yaboot (second/cfg.c); it has been cleaned up a bit in yaboot but is
still basically the same.

Interesting.  That explains a lot.


And this script
adds spaces and quotes around key/value pairs and let /dev/disk/by-*
names. This is not compatible with yaboot.

Which of those things is not compatible with yaboot?


It's not yaboot per-se that is incompatible. The extra quotes and spaces mess up the "ybin" script which is used in installing yaboot. My patch fixes that specific problem (and a couple of other specific problems) in a very non-general way. I'd recommend a complete rewrite of ybin (at least) if you want to make it work for general lilo-like syntax.



I also saw that you filled this bug:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=581173 which make me
think that your approach is to adapt every bootloader to this syntax.
[...]

I filed that bug because elilo.sh is not consistent with ELILO itself.

Ben.

--
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.


Reply to: