Added AST generation support for strace mutation

[ex0dus-0x]

This pull request adds the generation of a Pythonic abstract syntax tree from raw strace, making it ideal for CrashSimulator mutation.

Most of the work has been done through extending the Trace interface, but adding support for the to_ast() and from_ast() instance methods, as seen in this example:

from posix_omni_parser import Trace

trace = Trace.Trace("/path/to/trace", "/path/to/syscall_definitions.pickle")

# generate AST
mutable_ast = trace.to_ast()

######
# ... perform mutation to trace
######

# convert back to object
trace.from_ast(mutable_ast)

This is a snippet of the trace that could be produced as an AST:

{
    'name': 'read'
    'pid': '3172', 
    'args': [
        {'arg_num': 0, 'class': 'FileDescriptor', 'value': '3'}, 
        {'arg_num': 1, 'class': 'UnimplementedType', 'value': '""'}, 
        {'arg_num': 2, 'class': 'Hex', 'value': '1024'}
    ], 
   'type': 2,
   'return': 0, 
},

Features

• Extension of Trace object for AST generation and import/export
• Addition of ASTParser for actually generating the parse tree with support for Pythonic querying and mutation
• Extension of Syscall.py and parsing_classes.py for AST representation support.

[ex0dus-0x]

One way that I have realized this is broken is through __repr__ of the Syscall object.

# generate AST
# hack up and modify AST
# reimport AST
trace.from_ast(mutable_ast)

print trace.syscalls
# AttributeError!
[Traceback (most recent call last):
File "__main__.py", line 55, in <module>
  main()
File "__main__.py", line 52, in main
  print trace.syscalls
File "/home/testbed/posix-omni-parser/posix_omni_parser/Syscall.py", line 221, in __repr__                                                                                 
  + "RETURN:        " + str(self.ret) + "\n" \

This happens despite the fact that all self.ret values have been parsed correctly, and that isolated test cases were also used to see if this still occurs.

[pkmoore]

What needs to be fixed in order for this to be ready to merge?

[ex0dus-0x]

Possibly changes in how printing a trace's syscall attribute through __str__ and __repr__.

Right now, this is able to work for the purposes of hooking in for trace mutation (This might even be a method itself for writing trace anomalies).