Re: Bug#260512: swap is turned OFF after disk formatting!

[Joey Hess <joeyh@debian.org>]

Anton Zinoviev wrote:
> My only change in this release of partman was to repair the function
> enable_swap in definitions.sh.  The final swapon must have been broken
> somehow long ago.
> 
> #260746 (swap is not in /etc/fstab) is also not new.  I don't understand
> how this was possible but this bug seams to be there since version 22 of
> partman-basicfilesystems (11 Apr 2004 !).  If there was at least one bug
> report about this problem, I would fix it.

I worked out exactly what caused this all yesterday. You changed swap to
a method, but fstab.d/basic had this code:

            [ -f "$id/method" ] || continue
            [ -f "$id/acting_filesystem" ] || continue
            method=$(cat $id/method)
            if [ "$method" = swap ]; then
                echo "$path" none swap sw 0 0
            fi

Swap partitions no longer have an acting_filesystem file (maybe this is
a bug), and this made it not output the fstab entries. Also 
finish.d/mount_partitions no longer saw the swap entries, so it
neglected to call mount.d on the swap partitions to swapon them. I fixed
the bug yesterday and swap is again working well (although my other
changes to make it swapon as early as possible, and only swapoff when
really necessary introduced a new bug that I only fixed today).

> Sory about this.  I only want to say that most likely both problems are
> not caused by my recent changes in partman.

I'm quite sure that they were caused by making swap a method.

Also, Christian Perrier has reported a problem where partman hangs on
startup (at init.d/35something), it seems that parted_server has a pid
file but is not running. I've not gotten any more info to debug this but
it seems this is also a new problem since Christian does lots of tests
in vmware all the time.

I'm still not sure if we'll be able to include the new partman in the
next release, since we still seem to be finding new bugs in it.

-- 
see shy jo