We do not want to have all the stuff in one file.

Easier to maintain.

Do not print to stdout if not desired and make
it stop after a given number of iterations (if given).
Also allow to specify the starting set of paths.

So that we can store additional information with path

Well, not only loop headers, just on all  places to which
there is a DFS backedge.

A class that visits nodes or edges in the DFS order and calls
a given function on them (passes the information about node/edge
and DFS edge classification).

In that case, include all the error found along the path,
not only those that are found in the last step of the path.

That is, execute the prefix of the path and then do normal
fork instead of following the error edge.

type is the 2nd argument

So that we can use concrete domain.

Use the core abilities of the solver.

We do not want to make a copy of the state
(and if we do, we want to further use that copy -- that was the problem
in assume() method).

That is, do not add unnecessary constraints.

Reveals a bug in handling assume statement.

We must also add init paths when extending by more than 1 step.

We added a wrong path to the returnlist.

We changed the command line options

This argument is used only for symbolic execution.

Oh, and rename files to match the arguments.

Check that a non-constant assumption is SAT.
Without this check, we can continue executing an infeasible
path which than later breaks forking as neither the condition
not its negation is satisfiable.

We renamed the options.

Restruture files and options for -kind.