Seed random on policy init to make results deterministic (#466)

* seed random on policy init to make results deterministic

* document new random behavior

docs/gotchas.rst

@@ -30,4 +30,7 @@ Client code should use the :meth:`~manticore.Manticore.locked_context` API::
         global_context['flag1'] += ['a']
         global_context['flag2'] += ['b']
 
+"Random" Policy
+---------------
 
+The `random` policy, which is the manticore default, is not actually random and is instead deterministically seeded. This means that running the same analysis twice should return the same results (and get stuck in the same places).

manticore/core/executor.py

@@ -77,6 +77,7 @@ class Policy(object):
 class Random(Policy):
     def __init__(self, executor, *args, **kwargs):
         super(Random, self).__init__(executor, *args, **kwargs)
+        random.seed(1337) # For repeatable results
 
     def choice(self, state_ids):
         return random.choice(state_ids)