Bug#514967: Processed: submitter 514967, submitter 514976

[Colin Watson]

tags 514967 - fixed-upstream
tags 514976 - fixed-upstream
tags 533898 - fixed-upstream
tags 558422 - fixed-upstream
tags 584499 - fixed-upstream
tags 624263 - fixed-upstream
user phcoder at gmail.com
usertags 514967 not-upstream
usertags 514976 not-upstream
usertags 533898 not-upstream
usertags 558422 not-upstream
usertags 584499 not-upstream
usertags 624263 not-upstream
user grub2 at packages.debian.org
usertags 514967 grub-mkdevicemap
usertags 514976 grub-mkdevicemap
usertags 533898 grub-mkdevicemap
usertags 558422 grub-mkdevicemap
usertags 584499 grub-mkdevicemap
usertags 624263 grub-mkdevicemap
thanks

On Sat, Feb 04, 2012 at 01:20:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> tag 514967 fixed-upstream
> thanks
> >Bug #514967 [grub-common] grub-mkdevicelist does not cope with /boot on a disk>  #16
> >Changed Bug submitter to 'Brian May<brian at vpac.org>' from 'Chris Samuel<csamuel at vpac.org>'

So, this was fixed by removing grub-mkdevicemap.  In general I think
this is a good direction to be moving in; all the device hardcoding in
GRUB was always troublesome.

However, we have a lot of local scripts which depend on grub-mkdevicemap
in one way or another.  Of course some of these are trivial and could
easily be removed, but some aren't.  In particular, there's code in the
Debian postinst which offers a list of possible devices to install on,
using 'grub-mkdevicemap -m -' to find this list; grub-installer (part of
the Debian installer) uses grub-mkdevicemap to guess a reasonable
default boot device; in Ubuntu, lupin (part of Wubi) has a horrible hack
to look for wubildr on all partitions of BIOS-accessible disk devices,
which uses grub-mkdevicemap similarly; and so on.

Now, we could certainly go off and rewrite all these to use something
else.  It would probably end up being somewhat OS-specific: for example
on Linux we might use udevadm.  But really, this doesn't seem like a
desperately good use of time when the code already exists.  I understand
that there's probably very little interest in maintaining this code
upstream, so, for the time being, I'm resurrecting the last version of
it in a Debian-specific patch to 2.00.

I'm adjusting the tags on all affected bugs I could easily find to
reflect this.

-- 
Colin Watson                                       [cjwatson at debian.org]