1 from __future__ import unicode_literals
14 def _extract_tags(file_contents):
15 if file_contents[1:3] != b'WS':
17 'Not an SWF file; header is %r' % file_contents[:3])
18 if file_contents[:1] == b'C':
19 content = zlib.decompress(file_contents[8:])
21 raise NotImplementedError(
22 'Unsupported compression format %r' %
25 # Determine number of bits in framesize rectangle
26 framesize_nbits = struct.unpack('!B', content[:1])[0] >> 3
27 framesize_len = (5 + 4 * framesize_nbits + 7) // 8
29 pos = framesize_len + 2 + 2
30 while pos < len(content):
31 header16 = struct.unpack('<H', content[pos:pos + 2])[0]
33 tag_code = header16 >> 6
34 tag_len = header16 & 0x3f
36 tag_len = struct.unpack('<I', content[pos:pos + 4])[0]
38 assert pos + tag_len <= len(content), \
39 ('Tag %d ends at %d+%d - that\'s longer than the file (%d)'
40 % (tag_code, pos, tag_len, len(content)))
41 yield (tag_code, content[pos:pos + tag_len])
45 class _AVMClass_Object(object):
46 def __init__(self, avm_class):
47 self.avm_class = avm_class
50 return '%s#%x' % (self.avm_class.name, id(self))
53 class _ScopeDict(dict):
54 def __init__(self, avm_class):
55 super(_ScopeDict, self).__init__()
56 self.avm_class = avm_class
59 return '%s__Scope(%s)' % (
61 super(_ScopeDict, self).__repr__())
64 class _AVMClass(object):
65 def __init__(self, name_idx, name):
66 self.name_idx = name_idx
68 self.method_names = {}
71 self.method_pyfunctions = {}
73 self.variables = _ScopeDict(self)
75 def make_object(self):
76 return _AVMClass_Object(self)
79 return '_AVMClass(%s)' % (self.name)
81 def register_methods(self, methods):
82 self.method_names.update(methods.items())
83 self.method_idxs.update(dict(
85 for name, idx in methods.items()))
88 def _read_int(reader):
94 b = struct.unpack('<B', buf)[0]
95 res = res | ((b & 0x7f) << shift)
103 res = _read_int(reader)
104 assert res & 0xf0000000 == 0
110 v = _read_int(reader)
111 if v & 0x80000000 != 0:
112 v = - ((v ^ 0xffffffff) + 1)
119 last_byte = b'\xff' if (ord(bs[2:3]) >= 0x80) else b'\x00'
120 return struct.unpack('<i', bs + last_byte)[0]
123 def _read_string(reader):
125 resb = reader.read(slen)
126 assert len(resb) == slen
127 return resb.decode('utf-8')
130 def _read_bytes(count, reader):
132 resb = reader.read(count)
133 assert len(resb) == count
137 def _read_byte(reader):
138 resb = _read_bytes(1, reader=reader)
139 res = struct.unpack('<B', resb)[0]
143 class SWFInterpreter(object):
144 def __init__(self, file_contents):
146 for tag_code, tag in _extract_tags(file_contents)
148 p = code_tag.index(b'\0', 4) + 1
149 code_reader = io.BytesIO(code_tag[p:])
151 # Parse ABC (AVM2 ByteCode)
153 # Define a couple convenience methods
154 u30 = lambda *args: _u30(*args, reader=code_reader)
155 s32 = lambda *args: _s32(*args, reader=code_reader)
156 u32 = lambda *args: _u32(*args, reader=code_reader)
157 read_bytes = lambda *args: _read_bytes(*args, reader=code_reader)
158 read_byte = lambda *args: _read_byte(*args, reader=code_reader)
160 # minor_version + major_version
165 for _c in range(1, int_count):
168 for _c in range(1, uint_count):
171 read_bytes(max(0, (double_count - 1)) * 8)
173 self.constant_strings = ['']
174 for _c in range(1, string_count):
175 s = _read_string(code_reader)
176 self.constant_strings.append(s)
177 namespace_count = u30()
178 for _c in range(1, namespace_count):
182 for _c in range(1, ns_set_count):
184 for _c2 in range(count):
186 multiname_count = u30()
195 0x0e: 2, # MultinameA
196 0x1b: 1, # MultinameL
197 0x1c: 1, # MultinameLA
199 self.multinames = ['']
200 for _c in range(1, multiname_count):
202 assert kind in MULTINAME_SIZES, 'Invalid multiname kind %r' % kind
204 u30() # namespace_idx
206 self.multinames.append(self.constant_strings[name_idx])
208 self.multinames.append('[MULTINAME kind: %d]' % kind)
209 for _c2 in range(MULTINAME_SIZES[kind]):
214 MethodInfo = collections.namedtuple(
216 ['NEED_ARGUMENTS', 'NEED_REST'])
218 for method_id in range(method_count):
221 for _ in range(param_count):
223 u30() # name index (always 0 for youtube)
225 if flags & 0x08 != 0:
228 for c in range(option_count):
231 if flags & 0x80 != 0:
232 # Param names present
233 for _ in range(param_count):
235 mi = MethodInfo(flags & 0x01 != 0, flags & 0x04 != 0)
236 method_infos.append(mi)
239 metadata_count = u30()
240 for _c in range(metadata_count):
243 for _c2 in range(item_count):
247 def parse_traits_info():
248 trait_name_idx = u30()
249 kind_full = read_byte()
250 kind = kind_full & 0x0f
251 attrs = kind_full >> 4
253 if kind in [0x00, 0x06]: # Slot or Const
255 u30() # type_name_idx
259 elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter
262 methods[self.multinames[trait_name_idx]] = method_idx
263 elif kind == 0x04: # Class
266 elif kind == 0x05: # Function
269 methods[function_idx] = self.multinames[trait_name_idx]
271 raise ExtractorError('Unsupported trait kind %d' % kind)
273 if attrs & 0x4 != 0: # Metadata present
274 metadata_count = u30()
275 for _c3 in range(metadata_count):
276 u30() # metadata index
283 for class_id in range(class_count):
286 cname = self.multinames[name_idx]
287 avm_class = _AVMClass(name_idx, cname)
288 classes.append(avm_class)
290 u30() # super_name idx
292 if flags & 0x08 != 0: # Protected namespace is present
293 u30() # protected_ns_idx
295 for _c2 in range(intrf_count):
299 for _c2 in range(trait_count):
300 trait_methods = parse_traits_info()
301 avm_class.register_methods(trait_methods)
303 assert len(classes) == class_count
304 self._classes_by_name = dict((c.name, c) for c in classes)
306 for avm_class in classes:
309 for _c2 in range(trait_count):
310 trait_methods = parse_traits_info()
311 avm_class.register_methods(trait_methods)
315 for _c in range(script_count):
318 for _c2 in range(trait_count):
322 method_body_count = u30()
323 Method = collections.namedtuple('Method', ['code', 'local_count'])
324 for _c in range(method_body_count):
328 u30() # init_scope_depth
329 u30() # max_scope_depth
331 code = read_bytes(code_length)
332 for avm_class in classes:
333 if method_idx in avm_class.method_idxs:
334 m = Method(code, local_count)
335 avm_class.methods[avm_class.method_idxs[method_idx]] = m
336 exception_count = u30()
337 for _c2 in range(exception_count):
344 for _c2 in range(trait_count):
347 assert p + code_reader.tell() == len(code_tag)
349 def extract_class(self, class_name):
351 return self._classes_by_name[class_name]
353 raise ExtractorError('Class %r not found' % class_name)
355 def extract_function(self, avm_class, func_name):
356 print('Extracting %s.%s' % (avm_class.name, func_name))
357 if func_name in avm_class.method_pyfunctions:
358 return avm_class.method_pyfunctions[func_name]
359 if func_name in self._classes_by_name:
360 return self._classes_by_name[func_name].make_object()
361 if func_name not in avm_class.methods:
362 raise ExtractorError('Cannot find function %s.%s' % (
363 avm_class.name, func_name))
364 m = avm_class.methods[func_name]
368 coder = io.BytesIO(m.code)
369 s24 = lambda: _s24(coder)
370 u30 = lambda: _u30(coder)
372 print('Invoking %s.%s(%r)' % (avm_class.name, func_name, tuple(args)))
373 registers = [avm_class.variables] + list(args) + [None] * m.local_count
375 scopes = collections.deque([
376 self._classes_by_name, avm_class.variables])
378 opcode = _read_byte(coder)
379 print('opcode: %r, stack(%d): %r' % (opcode, len(stack), stack))
380 if opcode == 17: # iftrue
384 coder.seek(coder.tell() + offset)
385 elif opcode == 18: # iffalse
389 coder.seek(coder.tell() + offset)
390 elif opcode == 36: # pushbyte
391 v = _read_byte(coder)
393 elif opcode == 42: # dup
396 elif opcode == 44: # pushstring
398 stack.append(self.constant_strings[idx])
399 elif opcode == 48: # pushscope
400 new_scope = stack.pop()
401 scopes.append(new_scope)
402 elif opcode == 70: # callproperty
404 mname = self.multinames[index]
406 args = list(reversed(
407 [stack.pop() for _ in range(arg_count)]))
410 if isinstance(obj, _AVMClass_Object):
411 func = self.extract_function(obj.avm_class, mname)
415 elif isinstance(obj, _ScopeDict):
416 if mname in obj.avm_class.method_names:
417 func = self.extract_function(obj.avm_class, mname)
423 elif isinstance(obj, compat_str):
425 assert len(args) == 1
426 assert isinstance(args[0], compat_str)
430 res = obj.split(args[0])
433 elif isinstance(obj, list):
435 assert len(args) == 1
436 assert isinstance(args[0], int)
440 elif mname == 'join':
441 assert len(args) == 1
442 assert isinstance(args[0], compat_str)
443 res = args[0].join(obj)
446 raise NotImplementedError(
447 'Unsupported property %r on %r'
449 elif opcode == 72: # returnvalue
452 elif opcode == 74: # constructproperty
455 args = list(reversed(
456 [stack.pop() for _ in range(arg_count)]))
459 mname = self.multinames[index]
460 assert isinstance(obj, _AVMClass)
461 construct_method = self.extract_function(
463 # We do not actually call the constructor for now;
464 # we just pretend it does nothing
465 stack.append(obj.make_object())
466 elif opcode == 79: # callpropvoid
468 mname = self.multinames[index]
470 args = list(reversed(
471 [stack.pop() for _ in range(arg_count)]))
473 if mname == 'reverse':
474 assert isinstance(obj, list)
477 raise NotImplementedError(
478 'Unsupported (void) property %r on %r'
480 elif opcode == 86: # newarray
483 for i in range(arg_count):
484 arr.append(stack.pop())
487 elif opcode == 93: # findpropstrict
489 mname = self.multinames[index]
490 for s in reversed(scopes):
496 stack.append(res[mname])
497 elif opcode == 94: # findproperty
499 mname = self.multinames[index]
500 for s in reversed(scopes):
505 res = avm_class.variables
507 elif opcode == 96: # getlex
509 mname = self.multinames[index]
510 for s in reversed(scopes):
515 scope = avm_class.variables
516 # I cannot find where static variables are initialized
517 # so let's just return None
518 res = scope.get(mname)
520 elif opcode == 97: # setproperty
523 idx = self.multinames[index]
526 elif opcode == 98: # getlocal
528 stack.append(registers[index])
529 elif opcode == 99: # setlocal
532 registers[index] = value
533 elif opcode == 102: # getproperty
535 pname = self.multinames[index]
536 if pname == 'length':
538 assert isinstance(obj, list)
539 stack.append(len(obj))
540 else: # Assume attribute access
542 assert isinstance(idx, int)
544 assert isinstance(obj, list)
545 stack.append(obj[idx])
546 elif opcode == 115: # convert_
548 intvalue = int(value)
549 stack.append(intvalue)
550 elif opcode == 128: # coerce
552 elif opcode == 133: # coerce_s
553 assert isinstance(stack[-1], (type(None), compat_str))
554 elif opcode == 160: # add
557 res = value1 + value2
559 elif opcode == 161: # subtract
562 res = value1 - value2
564 elif opcode == 164: # modulo
567 res = value1 % value2
569 elif opcode == 175: # greaterequals
572 result = value1 >= value2
574 elif opcode == 208: # getlocal_0
575 stack.append(registers[0])
576 elif opcode == 209: # getlocal_1
577 stack.append(registers[1])
578 elif opcode == 210: # getlocal_2
579 stack.append(registers[2])
580 elif opcode == 211: # getlocal_3
581 stack.append(registers[3])
582 elif opcode == 212: # setlocal_0
583 registers[0] = stack.pop()
584 elif opcode == 213: # setlocal_1
585 registers[1] = stack.pop()
586 elif opcode == 214: # setlocal_2
587 registers[2] = stack.pop()
588 elif opcode == 215: # setlocal_3
589 registers[3] = stack.pop()
591 raise NotImplementedError(
592 'Unsupported opcode %d' % opcode)
594 avm_class.method_pyfunctions[func_name] = resfunc