Fixed in the main repository now, too. Thanks for reporting!

No problem, I've already fixed my copy (and now I get more interesting results;
maybe not better, but more interesting ;))

Let's use this to get some useful practice with the workflow once I'm back from
the conference. Patrik: we'll fix this next week.

It is as simple as Patrik wrote, one just needs to change the paramater of the
hill_climbing method to be a double instead of an integer.
Malte, I sent you an email asking you something regargind the teamproject's
repository. So when you answered, I'll either fix the bug by then so you can
merge it or you can just directly change it in the master repository yourself
(it's just the two lines of code).

I'll check it later today or at least tomorrow.

Silvan, can you look into this?

Found a bug in the iPDB implementation
(search/pdbs/pattern_generation_haslum.h/.cc):

The average cost over all operators (named average_operator_cost throughout) is
computed in the initialize() method, which passes it to the hill_climbing
method, which passes it on to sample_states, where it is used to compute an
estimate of plan length. The average is (of course) represented by a double,
everywhere except in the parameter to the hill_climbing method. This means the
average gets rounded down when it's passed through hill_climbing, which means
that when it is < 1, the value that reaches sample_states is 0, which has
undesirable consequences.