#!/usr/bin/env python3 from test import support import array import io import marshal import sys import unittest import os import types class HelperMixin: def helper(self, sample, *extra): new = marshal.loads(marshal.dumps(sample, *extra)) self.assertEqual(sample, new) try: with open(support.TESTFN, "wb") as f: marshal.dump(sample, f, *extra) with open(support.TESTFN, "rb") as f: new = marshal.load(f) self.assertEqual(sample, new) finally: support.unlink(support.TESTFN) class IntTestCase(unittest.TestCase, HelperMixin): def test_ints(self): # Test the full range of Python ints. n = sys.maxsize while n: for expected in (-n, n): self.helper(expected) n = n >> 1 def test_int64(self): # Simulate int marshaling on a 64-bit box. This is most interesting if # we're running the test on a 32-bit box, of course. def to_little_endian_string(value, nbytes): b = bytearray() for i in range(nbytes): b.append(value & 0xff) value >>= 8 return b maxint64 = (1 << 63) - 1 minint64 = -maxint64-1 for base in maxint64, minint64, -maxint64, -(minint64 >> 1): while base: s = b'I' + to_little_endian_string(base, 8) got = marshal.loads(s) self.assertEqual(base, got) if base == -1: # a fixed-point for shifting right 1 base = 0 else: base >>= 1 def test_bool(self): for b in (True, False): self.helper(b) class FloatTestCase(unittest.TestCase, HelperMixin): def test_floats(self): # Test a few floats small = 1e-25 n = sys.maxsize * 3.7e250 while n > small: for expected in (-n, n): self.helper(float(expected)) n /= 123.4567 f = 0.0 s = marshal.dumps(f, 2) got = marshal.loads(s) self.assertEqual(f, got) # and with version <= 1 (floats marshalled differently then) s = marshal.dumps(f, 1) got = marshal.loads(s) self.assertEqual(f, got) n = sys.maxsize * 3.7e-250 while n < small: for expected in (-n, n): f = float(expected) self.helper(f) self.helper(f, 1) n *= 123.4567 class StringTestCase(unittest.TestCase, HelperMixin): def test_unicode(self): for s in ["", "Andr\xe8 Previn", "abc", " "*10000]: self.helper(marshal.loads(marshal.dumps(s))) def test_string(self): for s in ["", "Andr\xe8 Previn", "abc", " "*10000]: self.helper(s) def test_bytes(self): for s in [b"", b"Andr\xe8 Previn", b"abc", b" "*10000]: self.helper(s) class ExceptionTestCase(unittest.TestCase): def test_exceptions(self): new = marshal.loads(marshal.dumps(StopIteration)) self.assertEqual(StopIteration, new) class CodeTestCase(unittest.TestCase): def test_code(self): co = ExceptionTestCase.test_exceptions.__code__ new = marshal.loads(marshal.dumps(co)) self.assertEqual(co, new) def test_many_codeobjects(self): # Issue2957: bad recursion count on code objects count = 5000 # more than MAX_MARSHAL_STACK_DEPTH codes = (ExceptionTestCase.test_exceptions.__code__,) * count marshal.loads(marshal.dumps(codes)) def test_different_filenames(self): co1 = compile("x", "f1", "exec") co2 = compile("y", "f2", "exec") co1, co2 = marshal.loads(marshal.dumps((co1, co2))) self.assertEqual(co1.co_filename, "f1") self.assertEqual(co2.co_filename, "f2") @support.cpython_only def test_same_filename_used(self): s = """def f(): pass\ndef g(): pass""" co = compile(s, "myfile", "exec") co = marshal.loads(marshal.dumps(co)) for obj in co.co_consts: if isinstance(obj, types.CodeType): self.assertIs(co.co_filename, obj.co_filename) class ContainerTestCase(unittest.TestCase, HelperMixin): d = {'astring': 'foo@bar.baz.spam', 'afloat': 7283.43, 'anint': 2**20, 'ashortlong': 2, 'alist': ['.zyx.41'], 'atuple': ('.zyx.41',)*10, 'aboolean': False, 'aunicode': "Andr\xe8 Previn" } def test_dict(self): self.helper(self.d) def test_list(self): self.helper(list(self.d.items())) def test_tuple(self): self.helper(tuple(self.d.keys())) def test_sets(self): for constructor in (set, frozenset): self.helper(constructor(self.d.keys())) class BufferTestCase(unittest.TestCase, HelperMixin): def test_bytearray(self): b = bytearray(b"abc") self.helper(b) new = marshal.loads(marshal.dumps(b)) self.assertEqual(type(new), bytes) def test_memoryview(self): b = memoryview(b"abc") self.helper(b) new = marshal.loads(marshal.dumps(b)) self.assertEqual(type(new), bytes) def test_array(self): a = array.array('B', b"abc") new = marshal.loads(marshal.dumps(a)) self.assertEqual(new, b"abc") class BugsTestCase(unittest.TestCase): def test_bug_5888452(self): # Simple-minded check for SF 588452: Debug build crashes marshal.dumps([128] * 1000) def test_patch_873224(self): self.assertRaises(Exception, marshal.loads, '0') self.assertRaises(Exception, marshal.loads, 'f') self.assertRaises(Exception, marshal.loads, marshal.dumps(2**65)[:-1]) def test_version_argument(self): # Python 2.4.0 crashes for any call to marshal.dumps(x, y) self.assertEqual(marshal.loads(marshal.dumps(5, 0)), 5) self.assertEqual(marshal.loads(marshal.dumps(5, 1)), 5) def test_fuzz(self): # simple test that it's at least not *totally* trivial to # crash from bad marshal data for c in [chr(i) for i in range(256)]: try: marshal.loads(c) except Exception: pass def test_loads_recursion(self): s = b'c' + (b'X' * 4*4) + b'{' * 2**20 self.assertRaises(ValueError, marshal.loads, s) def test_recursion_limit(self): # Create a deeply nested structure. head = last = [] # The max stack depth should match the value in Python/marshal.c. if os.name == 'nt' and hasattr(sys, 'gettotalrefcount'): MAX_MARSHAL_STACK_DEPTH = 1500 else: MAX_MARSHAL_STACK_DEPTH = 2000 for i in range(MAX_MARSHAL_STACK_DEPTH - 2): last.append([0]) last = last[-1] # Verify we don't blow out the stack with dumps/load. data = marshal.dumps(head) new_head = marshal.loads(data) # Don't use == to compare objects, it can exceed the recursion limit. self.assertEqual(len(new_head), len(head)) self.assertEqual(len(new_head[0]), len(head[0])) self.assertEqual(len(new_head[-1]), len(head[-1])) last.append([0]) self.assertRaises(ValueError, marshal.dumps, head) def test_exact_type_match(self): # Former bug: # >>> class Int(int): pass # >>> type(loads(dumps(Int()))) # for typ in (int, float, complex, tuple, list, dict, set, frozenset): # Note: str subclasses are not tested because they get handled # by marshal's routines for objects supporting the buffer API. subtyp = type('subtyp', (typ,), {}) self.assertRaises(ValueError, marshal.dumps, subtyp()) # Issue #1792 introduced a change in how marshal increases the size of its # internal buffer; this test ensures that the new code is exercised. def test_large_marshal(self): size = int(1e6) testString = 'abc' * size marshal.dumps(testString) def test_invalid_longs(self): # Issue #7019: marshal.loads shouldn't produce unnormalized PyLongs invalid_string = b'l\x02\x00\x00\x00\x00\x00\x00\x00' self.assertRaises(ValueError, marshal.loads, invalid_string) def test_multiple_dumps_and_loads(self): # Issue 12291: marshal.load() should be callable multiple times # with interleaved data written by non-marshal code # Adapted from a patch by Engelbert Gruber. data = (1, 'abc', b'def', 1.0, (2, 'a', ['b', b'c'])) for interleaved in (b'', b'0123'): ilen = len(interleaved) positions = [] try: with open(support.TESTFN, 'wb') as f: for d in data: marshal.dump(d, f) if ilen: f.write(interleaved) positions.append(f.tell()) with open(support.TESTFN, 'rb') as f: for i, d in enumerate(data): self.assertEqual(d, marshal.load(f)) if ilen: f.read(ilen) self.assertEqual(positions[i], f.tell()) finally: support.unlink(support.TESTFN) def test_loads_reject_unicode_strings(self): # Issue #14177: marshal.loads() should not accept unicode strings unicode_string = 'T' self.assertRaises(TypeError, marshal.loads, unicode_string) def test_bad_reader(self): class BadReader(io.BytesIO): def read(self, n=-1): b = super().read(n) if n is not None and n > 4: b += b' ' * 10**6 return b for value in (1.0, 1j, b'0123456789', '0123456789'): self.assertRaises(ValueError, marshal.load, BadReader(marshal.dumps(value))) LARGE_SIZE = 2**31 pointer_size = 8 if sys.maxsize > 0xFFFFFFFF else 4 class NullWriter: def write(self, s): pass @unittest.skipIf(LARGE_SIZE > sys.maxsize, "test cannot run on 32-bit systems") class LargeValuesTestCase(unittest.TestCase): def check_unmarshallable(self, data): self.assertRaises(ValueError, marshal.dump, data, NullWriter()) @support.bigmemtest(size=LARGE_SIZE, memuse=2, dry_run=False) def test_bytes(self, size): self.check_unmarshallable(b'x' * size) @support.bigmemtest(size=LARGE_SIZE, memuse=2, dry_run=False) def test_str(self, size): self.check_unmarshallable('x' * size) @support.bigmemtest(size=LARGE_SIZE, memuse=pointer_size + 1, dry_run=False) def test_tuple(self, size): self.check_unmarshallable((None,) * size) @support.bigmemtest(size=LARGE_SIZE, memuse=pointer_size + 1, dry_run=False) def test_list(self, size): self.check_unmarshallable([None] * size) @support.bigmemtest(size=LARGE_SIZE, memuse=pointer_size*12 + sys.getsizeof(LARGE_SIZE-1), dry_run=False) def test_set(self, size): self.check_unmarshallable(set(range(size))) @support.bigmemtest(size=LARGE_SIZE, memuse=pointer_size*12 + sys.getsizeof(LARGE_SIZE-1), dry_run=False) def test_frozenset(self, size): self.check_unmarshallable(frozenset(range(size))) @support.bigmemtest(size=LARGE_SIZE, memuse=2, dry_run=False) def test_bytearray(self, size): self.check_unmarshallable(bytearray(size)) def test_main(): support.run_unittest(IntTestCase, FloatTestCase, StringTestCase, CodeTestCase, ContainerTestCase, ExceptionTestCase, BufferTestCase, BugsTestCase, LargeValuesTestCase, ) if __name__ == "__main__": test_main()