kes Add rather primitive device resource deadlock detection in
the job queue handler. It detects the same read and write device
for migration and cancels the job.
kes Start adding a unique_dbid_hander() routine in migration to avoid
duplicate mediaids and duplicate jobids.
kes Add patch from bug #708 to permit relative paths such as ../xxx
in the restore tree routine.
kes Add Eric Bollengier's patch to reduce locking time in the SD after
despooling, and thus many jobs finish faster.
kes Correct locking order of reservations lock and device mutex in
reserve.c -- this should correct Arno's deadlock that occurred
when doing mount/unmount at the same time a job is reserving a
drive.