As far as I can tell, it can be closed.

It seems like this issue could be closed unless we want to use this issue to 
actually make changes to the code.

The OptionCaveats have changed slightly. To two subsubsections of the subsection
"setting the "cost type" for landmark heuristics" are merged.

The LAMA-FF synergy master follows the same rules for setting the cost type as
the landmark heuristic, so the behavior is the same as reported in msg5903.
The LAMA-FF synergy slave inherits the cost type set by the masters transform
option.

Roundup's URL regexp caught too much. This is the link:

http://www.fast-downward.org/OptionCaveats

When working on this, please also update
http://www.fast-downward.org/OptionCaveats, which is (mostly) about this issue.

Done. issue190 is closed.

Sounds good.

I think issue190 and this issue talk about the same problem. Therefore, I suggest 
referencing the content of issue190 here, which is just two links to msg1060 and 
msg1063 and closing issue190 in favor of this newer issue.

The cost type used by landmark ff synergy heuristic can be set with lm_cost_type
option of landmark heuristic or transform option of synergy declaration:

lm_ff_syn(lm_rhw(lm_cost_type=normal), admissible=false,
transform=adapt_costs(normal))

While for admissible=true Fast Downward uses the transform option and ignores
the lm_cost_type option, for admissible=false both options are considered.
Following example shows the different results of for different options:

openstacks-opt08-strips/p01.pddl
--heuristic "hlm,hff=lm_ff_syn(lm_rhw(lm_cost_type=cost_type), admissible=false,
transform=adapt_costs(cost_type))" 
--search "lazy_greedy([hff,hlm],preferred=[hff,hlm])"

lm_cost_type=cost_type transform=adapt(cost_type) Expansions
normal                 normal                            536
normal                 one                                41
one                    normal                             28
one                    one                                18

We wish to clarify the use of cost type with landmark ff synergy heuristic. If
the heuristic intends to use an unique cost type, we should provide a single
option for cost type in the long term. Otherwise, we should document how the
existing options interleave within the landmark ff synergy heuristic.