emitting the failure message, it should do so at =>3

Thanks to Jools Wills <jools@oxfordinspire.co.uk> for the patch.

Thanks to Jools Wills <jools@oxfordinspire.co.uk> for the patch.

Conflicts:
	handlers/rsync.in

This workarounds a bug in duplicity. Quoting the reporter (thanks, Jonathan
Dupart <jonathan@dupart.org>, for the detailed report and initial patch):

The duplicity backend is broken when the gpg ecryption is set up and the
system default locale is not english (or C).

The bug reports on duplicity are here :
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=565398
https://bugs.launchpad.net/duplicity/+bug/510625

Without this change, e.g. 0.5.20 would not get --extra-clean, while it should.
Granted, the new test is buggy too, e.g. 0.7.1 will get --extra-clean,
while it should not. Will report that to Redmine.

Alexander Mette authored 11 years ago and intrigeri committed 11 years ago

duplicity >=0.6.20 now correctly removes useless signature files and
therefore the cache directory doesn't grow infinitely any more. Using
--extra-clean removes all signatures except from the last chain and so
renders older chains unusable, don't use it by default any more.
See: https://bugs.launchpad.net/duplicity/+bug/519948

A fix for debian #677410 on commit 29dcdbae introduces a new issue on wheezy
machines where some rsync_options gets out of the rsync command quotes when
invoked by su.

As reported, running backupninja through `bash -x` shows that the `--delete`
option from the rsync command is not included in the single quote for the `su
-c` command:

    nice -n 0 su -c '/usr/bin/rsync -av' --delete '--recursive
    --delete-excluded    //home/ /tmp/test//home//'

This patch attempt to solve that issue by using the intermediary variable
$command.

add additional sys backup options to provide the ability to backup the MBR for every device found, and to backup the BIOS (if the flashrom program is installed, and the mainboard is supported)

previously the code was modeled off of the partition backup methodology, this resulted in the code only attempting to detect luks on actual devices, and not on partitions. The code would step through every disk on the system (sda, sdb, etc.) and do a 'cryptsetup isLuks' on that device, and then based on that output save the luks header for any device that had one. Unfortunately, in many cases, luks headers are found on the partitions of those disks. So we were failing to detect any luks partitions and thus failing to backup their luks headers.

this commit changes that, it now will look for luks headers on the devices themselves (for cases where a full disk is used without partitions), and on any partitions that are detected