Done.

My initial guess of what was going wrong in the Promela domains was wrong: I
remember that this separation into different rules was a problem for the
validator at some point, but apparently it isn't anymore. After briefly checking
the PDDL 2.2 description, I'm not sure if separated rules of this kind are legal
or illegal, so I left them unmodified.

The real problem was the use of a type called "number", which is reserved for
actual numeric variables in PDDL levels 2 or above. I renamed the type to "number_".

> This could e.g. be done by adding a hint into the PDDL files (in a comment)
> stating which PDDL file the plans should be validated against.
>
> I guess that's a separate issue, though.

See issue219.

In Pathways-NoNeg, the problem is that the predicate "not-chosen" is not defined.

This version of Pathways is not a proper IPC formulation; I produced it in a
manual "compile away negations" step, and I guess I was being sloppy. I can fix
this, but I think it would actually be more useful if there were an automated
way to automatically validate plans against the *original* (not No-Neg) domain
and problem, since this would also test the correctness of the reformulation.

This could e.g. be done by adding a hint into the PDDL files (in a comment)
stating which PDDL file the plans should be validated against.

I guess that's a separate issue, though.

The following domains cause problems:

philosophers, optical-telegraphs, pathways-noneg

Steps to reproduce: Find a plan for the first instance of these domains and run 
the validator on domain, problem and plan file.

Malte suggests for the first two domains definitions of the form


(:derived (blocked-trans ?p - process ?t - transition) ... )
(:derived (blocked-trans ?p - process ?t - transition) ... )

Should be combined to one definition.