The current revision in default correctly reports the error and exits with
EXIT_UNSUPPORTED (3).

Please test with the command-line Silvia mentions on a problem with axioms (e.g.
from psr-large?). If it exits with a clean error message, please close.

This issue seems to have a duplicate that is already fixed (issue380). Should we
close it or is there something else to it?

see also issue247.

I see where this error is coming from - many references to g_operators, 
sometimes from functions which might return an axiom id 
(get_operators_including_eff) - but I think this should also be assigned to 
Emil.

Regarding checking about axioms/conditional effects/non-unit cost actions/zero 
cost actions/... - I suggest we implement some sort of capabilities system 
(similar to weka), which would give us a convenient way of checking all of 
these. I'll open a separate issue for this.

I think I used

--search "astar(lmcount(lm_merged(lm_rhw(),lm_hm(m=1)),admissible=true),mpd=true)"

In this case it's the hm landmark generation that segfaults on axioms.
Of course it would be good if the hm landmarks would print an error message
rather than segfaulting. So maybe the title of this issue was not well chosen
and it should be "hm landmarks segfault on axioms".

But I think it would also be good though if there could be a check for axioms
(and a clean exit with error message) before any LM generation is even started.
Since the cost partitioning in the heuristic can't handle axioms, there's no
point in calculating landmarks at all, right? Is there a way that we can check
the capabilities of heuristics and other components like LM generations against
the requirements of the input, and print appropriate error messages right in the
beginning?

Emil is not on the tracker it seems. I can ask him if he'd like to join.

The first achievers are calculated in the LandmarksGraph - before the heuristic 
is created. What type of landmarks did you use here?
Can you attach the command line?

While M&S and lmcut abort with a clean error message ("can't handle
axioms") right at the start, the optimal landmark count heuristic
doesn't do this. There is an error message saying that the
cost partitioning can't handle axioms, but this message comes too late
-- you only get to see it if you actually _get_ to the part of the
code that does the cost partitioning. If you use h_m landmarks, you
get a segmentation fault before getting that far. The seg fault happens when
the code tries to calculate first achievers.