- def _parse_sig_swf(self, file_contents):
- if file_contents[1:3] != b'WS':
- raise ExtractorError(
- u'Not an SWF file; header is %r' % file_contents[:3])
- if file_contents[:1] == b'C':
- content = zlib.decompress(file_contents[8:])
- else:
- raise NotImplementedError(u'Unsupported compression format %r' %
- file_contents[:1])
-
- def extract_tags(content):
- pos = 0
- while pos < len(content):
- header16 = struct.unpack('<H', content[pos:pos+2])[0]
- pos += 2
- tag_code = header16 >> 6
- tag_len = header16 & 0x3f
- if tag_len == 0x3f:
- tag_len = struct.unpack('<I', content[pos:pos+4])[0]
- pos += 4
- assert pos+tag_len <= len(content)
- yield (tag_code, content[pos:pos+tag_len])
- pos += tag_len
-
- code_tag = next(tag
- for tag_code, tag in extract_tags(content)
- if tag_code == 82)
- p = code_tag.index(b'\0', 4) + 1
-
- # Parse ABC (AVM2 ByteCode)
- def read_int(data=None, pos=None):
- if hasattr(data, 'read'):
- assert pos is None
-
- res = 0
- shift = 0
- for _ in range(5):
- buf = data.read(1)
- assert len(buf) == 1
- b = struct.unpack('<B', buf)[0]
- res = res | ((b & 0x7f) << shift)
- if b & 0x80 == 0:
- break
- shift += 7
- return res
-
- if data is None:
- data = code_tag
- if pos is None:
- pos = p
- res = 0
- shift = 0
- for _ in range(5):
- b = struct.unpack('<B', data[pos:pos+1])[0]
- pos += 1
- res = res | ((b & 0x7f) << shift)
- if b & 0x80 == 0:
- break
- shift += 7
- return (res, pos)
- assert read_int(b'\x00', 0) == (0, 1)
- assert read_int(b'\x10', 0) == (16, 1)
- assert read_int(b'\x34', 0) == (0x34, 1)
- assert read_int(b'\xb4\x12', 0) == (0x12 * 0x80 + 0x34, 2)
- assert read_int(b'\xff\xff\xff\x00', 0) == (0x1fffff, 4)
-
- def u30(*args, **kwargs):
- res = read_int(*args, **kwargs)
- if isinstance(res, tuple):
- assert res[0] & 0xf0000000 == 0
- else:
- assert res & 0xf0000000 == 0
- return res
- u32 = read_int
-
- def s32(data=None, pos=None):
- v, pos = read_int(data, pos)
- if v & 0x80000000 != 0:
- v = - ((v ^ 0xffffffff) + 1)
- return (v, pos)
- assert s32(b'\xff\xff\xff\xff\x0f', 0) == (-1, 5)
-
- def string():
- slen, p = u30()
- return (code_tag[p:p+slen].decode('utf-8'), p + slen)
-
- def read_byte(data=None, pos=None):
- if data is None:
- data = code_tag
- if pos is None:
- pos = p
- res = struct.unpack('<B', data[pos:pos+1])[0]
- return (res, pos + 1)
-
- # minor_version + major_version
- p += 2 + 2
-
- # Constant pool
- int_count, p = u30()
- for _c in range(1, int_count):
- _, p = s32()
- uint_count, p = u30()
- for _c in range(1, uint_count):
- _, p = u32()
- double_count, p = u30()
- p += (double_count-1) * 8
- string_count, p = u30()
- constant_strings = [u'']
- for _c in range(1, string_count):
- s, p = string()
- constant_strings.append(s)
- namespace_count, p = u30()
- for _c in range(1, namespace_count):
- p += 1 # kind
- _, p = u30() # name
- ns_set_count, p = u30()
- for _c in range(1, ns_set_count):
- count, p = u30()
- for _c2 in range(count):
- _, p = u30()
- multiname_count, p = u30()
- MULTINAME_SIZES = {
- 0x07: 2, # QName
- 0x0d: 2, # QNameA
- 0x0f: 1, # RTQName
- 0x10: 1, # RTQNameA
- 0x11: 0, # RTQNameL
- 0x12: 0, # RTQNameLA
- 0x09: 2, # Multiname
- 0x0e: 2, # MultinameA
- 0x1b: 1, # MultinameL
- 0x1c: 1, # MultinameLA
- }
- multinames = [u'']
- for _c in range(1, multiname_count):
- kind, p = u30()
- assert kind in MULTINAME_SIZES, u'Invalid multiname kind %r' % kind
- if kind == 0x07:
- namespace_idx, p = u30()
- name_idx, p = u30()
- multinames.append(constant_strings[name_idx])
- else:
- multinames.append('[MULTINAME kind: %d]' % kind)
- for _c2 in range(MULTINAME_SIZES[kind]):
- _, p = u30()
-
- # Methods
- method_count, p = u30()
- MethodInfo = collections.namedtuple(
- 'MethodInfo',
- ['NEED_ARGUMENTS', 'NEED_REST'])
- method_infos = []
- for method_id in range(method_count):
- param_count, p = u30()
- _, p = u30() # return type
- for _ in range(param_count):
- _, p = u30() # param type
- _, p = u30() # name index (always 0 for youtube)
- flags, p = read_byte()
- if flags & 0x08 != 0:
- # Options present
- option_count, p = u30()
- for c in range(option_count):
- _, p = u30() # val
- p += 1 # kind
- if flags & 0x80 != 0:
- # Param names present
- for _ in range(param_count):
- _, p = u30() # param name
- mi = MethodInfo(flags & 0x01 != 0, flags & 0x04 != 0)
- method_infos.append(mi)
-
- # Metadata
- metadata_count, p = u30()
- for _c in range(metadata_count):
- _, p = u30() # name
- item_count, p = u30()
- for _c2 in range(item_count):
- _, p = u30() # key
- _, p = u30() # value
-
- def parse_traits_info(pos=None):
- if pos is None:
- pos = p
- trait_name_idx, pos = u30(pos=pos)
- kind_full, pos = read_byte(pos=pos)
- kind = kind_full & 0x0f
- attrs = kind_full >> 4
- methods = {}
- if kind in [0x00, 0x06]: # Slot or Const
- _, pos = u30(pos=pos) # Slot id
- type_name_idx, pos = u30(pos=pos)
- vindex, pos = u30(pos=pos)
- if vindex != 0:
- _, pos = read_byte(pos=pos) # vkind
- elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter
- _, pos = u30(pos=pos) # disp_id
- method_idx, pos = u30(pos=pos)
- methods[multinames[trait_name_idx]] = method_idx
- elif kind == 0x04: # Class
- _, pos = u30(pos=pos) # slot_id
- _, pos = u30(pos=pos) # classi
- elif kind == 0x05: # Function
- _, pos = u30(pos=pos) # slot_id
- function_idx, pos = u30(pos=pos)
- methods[function_idx] = multinames[trait_name_idx]
- else:
- raise ExtractorError(u'Unsupported trait kind %d' % kind)
-
- if attrs & 0x4 != 0: # Metadata present
- metadata_count, pos = u30(pos=pos)
- for _c3 in range(metadata_count):
- _, pos = u30(pos=pos)
-
- return (methods, pos)
-
- # Classes
- TARGET_CLASSNAME = u'SignatureDecipher'
- searched_idx = multinames.index(TARGET_CLASSNAME)
- searched_class_id = None
- class_count, p = u30()
- for class_id in range(class_count):
- name_idx, p = u30()
- if name_idx == searched_idx:
- # We found the class we're looking for!
- searched_class_id = class_id
- _, p = u30() # super_name idx
- flags, p = read_byte()
- if flags & 0x08 != 0: # Protected namespace is present
- protected_ns_idx, p = u30()
- intrf_count, p = u30()
- for _c2 in range(intrf_count):
- _, p = u30()
- _, p = u30() # iinit
- trait_count, p = u30()
- for _c2 in range(trait_count):
- _, p = parse_traits_info()
-
- if searched_class_id is None:
- raise ExtractorError(u'Target class %r not found' %
- TARGET_CLASSNAME)
-
- method_names = {}
- method_idxs = {}
- for class_id in range(class_count):
- _, p = u30() # cinit
- trait_count, p = u30()
- for _c2 in range(trait_count):
- trait_methods, p = parse_traits_info()
- if class_id == searched_class_id:
- method_names.update(trait_methods.items())
- method_idxs.update(dict(
- (idx, name)
- for name, idx in trait_methods.items()))
-
- # Scripts
- script_count, p = u30()
- for _c in range(script_count):
- _, p = u30() # init
- trait_count, p = u30()
- for _c2 in range(trait_count):
- _, p = parse_traits_info()
-
- # Method bodies
- method_body_count, p = u30()
- Method = collections.namedtuple('Method', ['code', 'local_count'])
- methods = {}
- for _c in range(method_body_count):
- method_idx, p = u30()
- max_stack, p = u30()
- local_count, p = u30()
- init_scope_depth, p = u30()
- max_scope_depth, p = u30()
- code_length, p = u30()
- if method_idx in method_idxs:
- m = Method(code_tag[p:p+code_length], local_count)
- methods[method_idxs[method_idx]] = m
- p += code_length
- exception_count, p = u30()
- for _c2 in range(exception_count):
- _, p = u30() # from
- _, p = u30() # to
- _, p = u30() # target
- _, p = u30() # exc_type
- _, p = u30() # var_name
- trait_count, p = u30()
- for _c2 in range(trait_count):
- _, p = parse_traits_info()
-
- assert p == len(code_tag)
- assert len(methods) == len(method_idxs)
-
- method_pyfunctions = {}
-
- def extract_function(func_name):
- if func_name in method_pyfunctions:
- return method_pyfunctions[func_name]
- if func_name not in methods:
- raise ExtractorError(u'Cannot find function %r' % func_name)
- m = methods[func_name]
-
- def resfunc(args):
- print('Entering function %s(%r)' % (func_name, args))
- registers = ['(this)'] + list(args) + [None] * m.local_count
- stack = []
- coder = io.BytesIO(m.code)
- while True:
- opcode = struct.unpack('!B', coder.read(1))[0]
- if opcode == 208: # getlocal_0
- stack.append(registers[0])
- elif opcode == 209: # getlocal_1
- stack.append(registers[1])
- elif opcode == 210: # getlocal_2
- stack.append(registers[2])
- elif opcode == 36: # pushbyte
- v = struct.unpack('!B', coder.read(1))[0]
- stack.append(v)
- elif opcode == 44: # pushstring
- idx = u30(coder)
- stack.append(constant_strings[idx])
- elif opcode == 48: # pushscope
- # We don't implement the scope register, so we'll just
- # ignore the popped value
- stack.pop()
- elif opcode == 70: # callproperty
- index = u30(coder)
- mname = multinames[index]
- arg_count = u30(coder)
- args = list(reversed(
- [stack.pop() for _ in range(arg_count)]))
- obj = stack.pop()
- if mname == u'split':
- assert len(args) == 1
- assert isinstance(args[0], compat_str)
- assert isinstance(obj, compat_str)
- if args[0] == u'':
- res = list(obj)
- else:
- res = obj.split(args[0])
- stack.append(res)
- elif mname in method_pyfunctions:
- stack.append(method_pyfunctions[mname](args))
- else:
- raise NotImplementedError(
- u'Unsupported property %r on %r'
- % (mname, obj))
- elif opcode == 93: # findpropstrict
- index = u30(coder)
- mname = multinames[index]
- res = extract_function(mname)
- stack.append(res)
- elif opcode == 97: # setproperty
- index = u30(coder)
- value = stack.pop()
- idx = stack.pop()
- obj = stack.pop()
- assert isinstance(obj, list)
- assert isinstance(idx, int)
- obj[idx] = value
- elif opcode == 98: # getlocal
- index = u30(coder)
- stack.append(registers[index])
- elif opcode == 99: # setlocal
- index = u30(coder)
- value = stack.pop()
- registers[index] = value
- elif opcode == 102: # getproperty
- index = u30(coder)
- pname = multinames[index]
- if pname == u'length':
- obj = stack.pop()
- assert isinstance(obj, list)
- stack.append(len(obj))
- else: # Assume attribute access
- idx = stack.pop()
- assert isinstance(idx, int)
- obj = stack.pop()
- assert isinstance(obj, list)
- stack.append(obj[idx])
- elif opcode == 128: # coerce
- _ = u30(coder)
- elif opcode == 133: # coerce_s
- assert isinstance(stack[-1], (type(None), compat_str))
- elif opcode == 164: # modulo
- value2 = stack.pop()
- value1 = stack.pop()
- res = value1 % value2
- stack.append(res)
- elif opcode == 214: # setlocal_2
- registers[2] = stack.pop()
- elif opcode == 215: # setlocal_3
- registers[3] = stack.pop()
- else:
- raise NotImplementedError(
- u'Unsupported opcode %d' % opcode)