Merged.

Works for me. :-)

Experiments used for the paper:
http://ai.cs.unibas.ch/_tmp_files/sieverss/2016-02-24-dfp-abs.html

Experiments from this issue branch:
http://ai.cs.unibas.ch/_tmp_files/sieverss/issue656-v1-new-configs.html

Comparing coverage, we never lose a task, and we gain a few for some configurations.

The behavior of "default" DFP did not change:

http://ai.cs.unibas.ch/_tmp_files/sieverss/issue656-v1-issue656-base-issue656-v1-compare.html

(Experiments to confirm the results from the paper are still ongoing.)

Ok then. I am very sure to do nothing wrong here, because the change only
affects MergeDFP (and has much less impact than the change from issue655, which
changed the interface of shrink strategies). Furthermore, with issue644
(re-design of merge strategies), this class will no longer exist in this form
anyway. So it will probably be better if you have some energy left over for
having a look at issue644 when it comes to code reviewing :-)

OK, I can't find the energy to look at it in much detail, so I'll just assume
that you know what you're doing. :-) From my side, feel free to merge.

There it is: https://bitbucket.org/SilvanS/fd-dev/pull-requests/15/issue656/diff

As soon as I extracted the required changes to an issue branch (not done yet).

Do you have a link for a pull request (or other viewable diff)?

I'd like to merge the changes that add the "ordering of transition systems"
options (ICAPS 16 paper) to the DFP merge strategy. The code change is small,
and integrating this before starting to work on the huge refactoring of merge
strategies (issue644) would allow to make sure that the behavior of these
options remains the same when changing MergeDFP to use scoring functions.