Proper reporting for this has been added; all duplicate objects are now
mentioned on the command line and inputs with duplicates are not processed further.
(This showed that there are further duplicates in the input, but apparently
these didn't confuse the planner to the extent of rendering the problem unsolvable.)

Forbidding duplicate declarations seems fine to me. It certainly seems consistent 
with the PDDL specification.

Thank you very much for catching this!  Catching this in the translator would be 
very helpful.

The question also extends to the case where we have multiple definitions of the
same object in the same block (:objects or :constants).

Thanks for the report!

I would argue that the file is invalid -- the object "m_free" is defined twice,
once as a constant in "trap.pddl" and then again as an object in "problem.pddl".
The planner gets confused about the identity of these two, and hence does not
realize that the various occurrences of "m_free" all refer to the same object.
If you get rid of the second definition of m_free in problem.pddl, the problem
is solved.

Of course we need better error reporting here, and also the semantics of this is
arguable. I'd be tempted to generally disallow duplicate definitions of objects
since otherwise we'd have to think about what should happen

1) if one of them has a type attached and the other happens
2) if both have a type and they are the same
3) if both have a type and one is a supertype of the other
4) if both have a type and they are incomparable

What do you think?

After more experiments, I find that many variations of this domain and problem
are similarly spuriously treated as unsolvable by FD.

I have a domain and problem with a trivial one-step plan solution, that FF is
able to find (and validate accepts), but that FD is unable to find, configured
as FF or LAMA.  Worse, FD assures me that the plan is not solvable:
ff -o trap.pddl -f problem.pddl

ff: parsing domain file
domain 'TRAP' defined
 ... done.
ff: parsing problem file
problem 'TEST' defined
 ... done.



Cueing down from goal distance:    1 into depth [1]
                                   0            

ff: found legal plan as follows

step    0: GO_TO_HOST ANDY BRAVO
     

time spent:   32.99 seconds instantiating 204597 easy, 1186460 hard action templates
               7.85 seconds reachability analysis, yielding 1448 facts and 52831
actions
               0.45 seconds creating final representation with 1101 relevant facts
               0.76 seconds building connectivity graph
               0.06 seconds searching, evaluating 2 states, to a max depth of 1
              42.11 seconds total time

 fast-downward trap.pddl problem.pddl
Parsing... [0.050s CPU, 0.035s wall-clock]
Instantiating...
Normalizing task... [0.017s CPU, 0.006s wall-clock]
Generating Datalog program... [0.000s CPU, 0.007s wall-clock]
Normalizing Datalog program...
Normalizing Datalog program: [0.150s CPU, 0.087s wall-clock]
Preparing model... [0.083s CPU, 0.040s wall-clock]
Generated 961 rules.
Computing model... [1.333s CPU, 0.806s wall-clock]
8064 relevant atoms
23918 auxiliary atoms
31982 final queue length
50760 total queue pushes
Completing instantiation... [2.300s CPU, 1.385s wall-clock]
Instantiating: [3.933s CPU, 2.358s wall-clock]
Computing fact groups...
Finding invariants...
159 initial candidates
Finding invariants: [0.133s CPU, 0.083s wall-clock]
Checking invariant weight... [0.000s CPU, 0.001s wall-clock]
Instantiating groups... [0.017s CPU, 0.010s wall-clock]
Collecting mutex groups... [0.000s CPU, 0.001s wall-clock]
Choosing groups...
1067 uncovered facts
Choosing groups: [0.000s CPU, 0.001s wall-clock]
Building translation key... [0.017s CPU, 0.004s wall-clock]
Computing fact groups: [0.167s CPU, 0.099s wall-clock]
Building STRIPS to SAS dictionary... [0.000s CPU, 0.002s wall-clock]
Building dictionary for full mutex groups... [0.000s CPU, 0.002s wall-clock]
Building mutex information...
Building mutex information: [0.000s CPU, 0.003s wall-clock]
Translating task...
Processing axioms...
Simplifying axioms... [0.000s CPU, 0.000s wall-clock]
Processing axioms: [0.067s CPU, 0.038s wall-clock]
Translating task: [0.133s CPU, 0.075s wall-clock]
0 implied effects removed
0 effect conditions simplified
0 implied preconditions added
Detecting unreachable propositions...
Simplified to trivially false goal! Generating unsolvable task...
Detecting unreachable propositions: [0.017s CPU, 0.011s wall-clock]
Translator variables: 1
Translator derived variables: 0
Translator facts: 2
Translator mutex groups: 0
Translator total mutex groups size: 0
Translator operators: 0
Translator task size: 4
Writing output... [0.000s CPU, 0.000s wall-clock]
Done! [4.367s CPU, 2.624s wall-clock]
Running preprocessor with timelimit of 298
Building causal graph...
The causal graph is acyclic.
1 variables of 1 necessary
0 of 0 mutex groups necessary.
0 of 0 operators necessary.
0 of 0 axiom rules necessary.
Building domain transition graphs...
solveable in poly time 0
Building successor generator...
Preprocessor facts: 2
Preprocessor derived variables: 0
Preprocessor task size: 4
Writing output...
done

Running search with timelimit of 296
Dispatcher selected state size 1.
This is a unit task.
Simplifying transitions... done!
Initializing Exploration...
Generating landmarks using the RPG/SAS+ approach
approx. reasonable orders
approx. obedient reasonable orders
Removed 0 reasonable or obedient reasonable orders
Landmarks generation time: 0.000119973s
Discovered 2 landmarks, of which 0 are disjunctive and 0 are conjunctive 
1 edges
Initializing LAMA-FF Synergy Object
Initializing landmarks count heuristic...
1 initial landmarks, 1 goal landmarks
Conducting lazy best first search, (real) bound = 2147483647
Completely explored state space -- no solution!
Actual search time: 0s [t=0s]
Expanded 0 state(s).
Reopened 0 state(s).
Evaluated 1 state(s).
Evaluations: 2
Generated 0 state(s).
Dead ends: 1 state(s).
Search time: 0s
Total time: 0s
Peak memory: 601916 KB