| import sys |
| import os |
| import io |
| import shutil |
| import io |
| from hashlib import md5 |
| import errno |
| |
| import unittest |
| import tarfile |
| |
| from test import support |
| |
| # Check for our compression modules. |
| try: |
| import gzip |
| gzip.GzipFile |
| except (ImportError, AttributeError): |
| gzip = None |
| try: |
| import bz2 |
| except ImportError: |
| bz2 = None |
| |
| def md5sum(data): |
| return md5(data).hexdigest() |
| |
| TEMPDIR = os.path.abspath(support.TESTFN) |
| tarname = support.findfile("testtar.tar") |
| gzipname = os.path.join(TEMPDIR, "testtar.tar.gz") |
| bz2name = os.path.join(TEMPDIR, "testtar.tar.bz2") |
| tmpname = os.path.join(TEMPDIR, "tmp.tar") |
| |
| md5_regtype = "65f477c818ad9e15f7feab0c6d37742f" |
| md5_sparse = "a54fbc4ca4f4399a90e1b27164012fc6" |
| |
| |
| class ReadTest(unittest.TestCase): |
| |
| tarname = tarname |
| mode = "r:" |
| |
| def setUp(self): |
| self.tar = tarfile.open(self.tarname, mode=self.mode, encoding="iso8859-1") |
| |
| def tearDown(self): |
| self.tar.close() |
| |
| |
| class UstarReadTest(ReadTest): |
| |
| def test_fileobj_regular_file(self): |
| tarinfo = self.tar.getmember("ustar/regtype") |
| fobj = self.tar.extractfile(tarinfo) |
| data = fobj.read() |
| self.assertTrue((len(data), md5sum(data)) == (tarinfo.size, md5_regtype), |
| "regular file extraction failed") |
| |
| def test_fileobj_readlines(self): |
| self.tar.extract("ustar/regtype", TEMPDIR) |
| tarinfo = self.tar.getmember("ustar/regtype") |
| fobj1 = open(os.path.join(TEMPDIR, "ustar/regtype"), "r") |
| fobj2 = io.TextIOWrapper(self.tar.extractfile(tarinfo)) |
| |
| lines1 = fobj1.readlines() |
| lines2 = fobj2.readlines() |
| self.assertTrue(lines1 == lines2, |
| "fileobj.readlines() failed") |
| self.assertTrue(len(lines2) == 114, |
| "fileobj.readlines() failed") |
| self.assertTrue(lines2[83] == |
| "I will gladly admit that Python is not the fastest running scripting language.\n", |
| "fileobj.readlines() failed") |
| |
| def test_fileobj_iter(self): |
| self.tar.extract("ustar/regtype", TEMPDIR) |
| tarinfo = self.tar.getmember("ustar/regtype") |
| fobj1 = open(os.path.join(TEMPDIR, "ustar/regtype"), "rU") |
| fobj2 = self.tar.extractfile(tarinfo) |
| lines1 = fobj1.readlines() |
| lines2 = list(io.TextIOWrapper(fobj2)) |
| self.assertTrue(lines1 == lines2, |
| "fileobj.__iter__() failed") |
| |
| def test_fileobj_seek(self): |
| self.tar.extract("ustar/regtype", TEMPDIR) |
| fobj = open(os.path.join(TEMPDIR, "ustar/regtype"), "rb") |
| data = fobj.read() |
| fobj.close() |
| |
| tarinfo = self.tar.getmember("ustar/regtype") |
| fobj = self.tar.extractfile(tarinfo) |
| |
| text = fobj.read() |
| fobj.seek(0) |
| self.assertEqual(0, fobj.tell(), |
| "seek() to file's start failed") |
| fobj.seek(2048, 0) |
| self.assertEqual(2048, fobj.tell(), |
| "seek() to absolute position failed") |
| fobj.seek(-1024, 1) |
| self.assertEqual(1024, fobj.tell(), |
| "seek() to negative relative position failed") |
| fobj.seek(1024, 1) |
| self.assertEqual(2048, fobj.tell(), |
| "seek() to positive relative position failed") |
| s = fobj.read(10) |
| self.assertTrue(s == data[2048:2058], |
| "read() after seek failed") |
| fobj.seek(0, 2) |
| self.assertEqual(tarinfo.size, fobj.tell(), |
| "seek() to file's end failed") |
| self.assertTrue(fobj.read() == b"", |
| "read() at file's end did not return empty string") |
| fobj.seek(-tarinfo.size, 2) |
| self.assertEqual(0, fobj.tell(), |
| "relative seek() to file's end failed") |
| fobj.seek(512) |
| s1 = fobj.readlines() |
| fobj.seek(512) |
| s2 = fobj.readlines() |
| self.assertTrue(s1 == s2, |
| "readlines() after seek failed") |
| fobj.seek(0) |
| self.assertEqual(len(fobj.readline()), fobj.tell(), |
| "tell() after readline() failed") |
| fobj.seek(512) |
| self.assertTrue(len(fobj.readline()) + 512 == fobj.tell(), |
| "tell() after seek() and readline() failed") |
| fobj.seek(0) |
| line = fobj.readline() |
| self.assertEqual(fobj.read(), data[len(line):], |
| "read() after readline() failed") |
| fobj.close() |
| |
| |
| class CommonReadTest(ReadTest): |
| |
| def test_empty_tarfile(self): |
| # Test for issue6123: Allow opening empty archives. |
| # This test checks if tarfile.open() is able to open an empty tar |
| # archive successfully. Note that an empty tar archive is not the |
| # same as an empty file! |
| tarfile.open(tmpname, self.mode.replace("r", "w")).close() |
| try: |
| tar = tarfile.open(tmpname, self.mode) |
| tar.getnames() |
| except tarfile.ReadError: |
| self.fail("tarfile.open() failed on empty archive") |
| self.assertListEqual(tar.getmembers(), []) |
| |
| def test_null_tarfile(self): |
| # Test for issue6123: Allow opening empty archives. |
| # This test guarantees that tarfile.open() does not treat an empty |
| # file as an empty tar archive. |
| open(tmpname, "wb").close() |
| self.assertRaises(tarfile.ReadError, tarfile.open, tmpname, self.mode) |
| self.assertRaises(tarfile.ReadError, tarfile.open, tmpname) |
| |
| def test_ignore_zeros(self): |
| # Test TarFile's ignore_zeros option. |
| if self.mode.endswith(":gz"): |
| _open = gzip.GzipFile |
| elif self.mode.endswith(":bz2"): |
| _open = bz2.BZ2File |
| else: |
| _open = open |
| |
| for char in (b'\0', b'a'): |
| # Test if EOFHeaderError ('\0') and InvalidHeaderError ('a') |
| # are ignored correctly. |
| fobj = _open(tmpname, "wb") |
| fobj.write(char * 1024) |
| fobj.write(tarfile.TarInfo("foo").tobuf()) |
| fobj.close() |
| |
| tar = tarfile.open(tmpname, mode="r", ignore_zeros=True) |
| self.assertListEqual(tar.getnames(), ["foo"], |
| "ignore_zeros=True should have skipped the %r-blocks" % char) |
| tar.close() |
| |
| |
| class MiscReadTest(CommonReadTest): |
| |
| def test_no_name_argument(self): |
| fobj = open(self.tarname, "rb") |
| tar = tarfile.open(fileobj=fobj, mode=self.mode) |
| self.assertEqual(tar.name, os.path.abspath(fobj.name)) |
| |
| def test_no_name_attribute(self): |
| data = open(self.tarname, "rb").read() |
| fobj = io.BytesIO(data) |
| self.assertRaises(AttributeError, getattr, fobj, "name") |
| tar = tarfile.open(fileobj=fobj, mode=self.mode) |
| self.assertEqual(tar.name, None) |
| |
| def test_empty_name_attribute(self): |
| data = open(self.tarname, "rb").read() |
| fobj = io.BytesIO(data) |
| fobj.name = "" |
| tar = tarfile.open(fileobj=fobj, mode=self.mode) |
| self.assertEqual(tar.name, None) |
| |
| def test_fileobj_with_offset(self): |
| # Skip the first member and store values from the second member |
| # of the testtar. |
| tar = tarfile.open(self.tarname, mode=self.mode) |
| tar.next() |
| t = tar.next() |
| name = t.name |
| offset = t.offset |
| data = tar.extractfile(t).read() |
| tar.close() |
| |
| # Open the testtar and seek to the offset of the second member. |
| if self.mode.endswith(":gz"): |
| _open = gzip.GzipFile |
| elif self.mode.endswith(":bz2"): |
| _open = bz2.BZ2File |
| else: |
| _open = open |
| fobj = _open(self.tarname, "rb") |
| fobj.seek(offset) |
| |
| # Test if the tarfile starts with the second member. |
| tar = tar.open(self.tarname, mode="r:", fileobj=fobj) |
| t = tar.next() |
| self.assertEqual(t.name, name) |
| # Read to the end of fileobj and test if seeking back to the |
| # beginning works. |
| tar.getmembers() |
| self.assertEqual(tar.extractfile(t).read(), data, |
| "seek back did not work") |
| tar.close() |
| |
| def test_fail_comp(self): |
| # For Gzip and Bz2 Tests: fail with a ReadError on an uncompressed file. |
| if self.mode == "r:": |
| return |
| self.assertRaises(tarfile.ReadError, tarfile.open, tarname, self.mode) |
| fobj = open(tarname, "rb") |
| self.assertRaises(tarfile.ReadError, tarfile.open, fileobj=fobj, mode=self.mode) |
| |
| def test_v7_dirtype(self): |
| # Test old style dirtype member (bug #1336623): |
| # Old V7 tars create directory members using an AREGTYPE |
| # header with a "/" appended to the filename field. |
| tarinfo = self.tar.getmember("misc/dirtype-old-v7") |
| self.assertTrue(tarinfo.type == tarfile.DIRTYPE, |
| "v7 dirtype failed") |
| |
| def test_xstar_type(self): |
| # The xstar format stores extra atime and ctime fields inside the |
| # space reserved for the prefix field. The prefix field must be |
| # ignored in this case, otherwise it will mess up the name. |
| try: |
| self.tar.getmember("misc/regtype-xstar") |
| except KeyError: |
| self.fail("failed to find misc/regtype-xstar (mangled prefix?)") |
| |
| def test_check_members(self): |
| for tarinfo in self.tar: |
| self.assertTrue(int(tarinfo.mtime) == 0o7606136617, |
| "wrong mtime for %s" % tarinfo.name) |
| if not tarinfo.name.startswith("ustar/"): |
| continue |
| self.assertTrue(tarinfo.uname == "tarfile", |
| "wrong uname for %s" % tarinfo.name) |
| |
| def test_find_members(self): |
| self.assertTrue(self.tar.getmembers()[-1].name == "misc/eof", |
| "could not find all members") |
| |
| def test_extract_hardlink(self): |
| # Test hardlink extraction (e.g. bug #857297). |
| tar = tarfile.open(tarname, errorlevel=1, encoding="iso8859-1") |
| |
| tar.extract("ustar/regtype", TEMPDIR) |
| try: |
| tar.extract("ustar/lnktype", TEMPDIR) |
| except EnvironmentError as e: |
| if e.errno == errno.ENOENT: |
| self.fail("hardlink not extracted properly") |
| |
| data = open(os.path.join(TEMPDIR, "ustar/lnktype"), "rb").read() |
| self.assertEqual(md5sum(data), md5_regtype) |
| |
| try: |
| tar.extract("ustar/symtype", TEMPDIR) |
| except EnvironmentError as e: |
| if e.errno == errno.ENOENT: |
| self.fail("symlink not extracted properly") |
| |
| data = open(os.path.join(TEMPDIR, "ustar/symtype"), "rb").read() |
| self.assertEqual(md5sum(data), md5_regtype) |
| |
| def test_extractall(self): |
| # Test if extractall() correctly restores directory permissions |
| # and times (see issue1735). |
| tar = tarfile.open(tarname, encoding="iso8859-1") |
| directories = [t for t in tar if t.isdir()] |
| tar.extractall(TEMPDIR, directories) |
| for tarinfo in directories: |
| path = os.path.join(TEMPDIR, tarinfo.name) |
| if sys.platform != "win32": |
| # Win32 has no support for fine grained permissions. |
| self.assertEqual(tarinfo.mode & 0o777, os.stat(path).st_mode & 0o777) |
| self.assertEqual(tarinfo.mtime, os.path.getmtime(path)) |
| tar.close() |
| |
| def test_init_close_fobj(self): |
| # Issue #7341: Close the internal file object in the TarFile |
| # constructor in case of an error. For the test we rely on |
| # the fact that opening an empty file raises a ReadError. |
| empty = os.path.join(TEMPDIR, "empty") |
| open(empty, "wb").write(b"") |
| |
| try: |
| tar = object.__new__(tarfile.TarFile) |
| try: |
| tar.__init__(empty) |
| except tarfile.ReadError: |
| self.assertTrue(tar.fileobj.closed) |
| else: |
| self.fail("ReadError not raised") |
| finally: |
| os.remove(empty) |
| |
| |
| class StreamReadTest(CommonReadTest): |
| |
| mode="r|" |
| |
| def test_fileobj_regular_file(self): |
| tarinfo = self.tar.next() # get "regtype" (can't use getmember) |
| fobj = self.tar.extractfile(tarinfo) |
| data = fobj.read() |
| self.assertTrue((len(data), md5sum(data)) == (tarinfo.size, md5_regtype), |
| "regular file extraction failed") |
| |
| def test_provoke_stream_error(self): |
| tarinfos = self.tar.getmembers() |
| f = self.tar.extractfile(tarinfos[0]) # read the first member |
| self.assertRaises(tarfile.StreamError, f.read) |
| |
| def test_compare_members(self): |
| tar1 = tarfile.open(tarname, encoding="iso8859-1") |
| tar2 = self.tar |
| |
| while True: |
| t1 = tar1.next() |
| t2 = tar2.next() |
| if t1 is None: |
| break |
| self.assertTrue(t2 is not None, "stream.next() failed.") |
| |
| if t2.islnk() or t2.issym(): |
| self.assertRaises(tarfile.StreamError, tar2.extractfile, t2) |
| continue |
| |
| v1 = tar1.extractfile(t1) |
| v2 = tar2.extractfile(t2) |
| if v1 is None: |
| continue |
| self.assertTrue(v2 is not None, "stream.extractfile() failed") |
| self.assertEqual(v1.read(), v2.read(), "stream extraction failed") |
| |
| tar1.close() |
| |
| |
| class DetectReadTest(unittest.TestCase): |
| |
| def _testfunc_file(self, name, mode): |
| try: |
| tarfile.open(name, mode) |
| except tarfile.ReadError as e: |
| self.fail() |
| |
| def _testfunc_fileobj(self, name, mode): |
| try: |
| tarfile.open(name, mode, fileobj=open(name, "rb")) |
| except tarfile.ReadError as e: |
| self.fail() |
| |
| def _test_modes(self, testfunc): |
| testfunc(tarname, "r") |
| testfunc(tarname, "r:") |
| testfunc(tarname, "r:*") |
| testfunc(tarname, "r|") |
| testfunc(tarname, "r|*") |
| |
| if gzip: |
| self.assertRaises(tarfile.ReadError, tarfile.open, tarname, mode="r:gz") |
| self.assertRaises(tarfile.ReadError, tarfile.open, tarname, mode="r|gz") |
| self.assertRaises(tarfile.ReadError, tarfile.open, gzipname, mode="r:") |
| self.assertRaises(tarfile.ReadError, tarfile.open, gzipname, mode="r|") |
| |
| testfunc(gzipname, "r") |
| testfunc(gzipname, "r:*") |
| testfunc(gzipname, "r:gz") |
| testfunc(gzipname, "r|*") |
| testfunc(gzipname, "r|gz") |
| |
| if bz2: |
| self.assertRaises(tarfile.ReadError, tarfile.open, tarname, mode="r:bz2") |
| self.assertRaises(tarfile.ReadError, tarfile.open, tarname, mode="r|bz2") |
| self.assertRaises(tarfile.ReadError, tarfile.open, bz2name, mode="r:") |
| self.assertRaises(tarfile.ReadError, tarfile.open, bz2name, mode="r|") |
| |
| testfunc(bz2name, "r") |
| testfunc(bz2name, "r:*") |
| testfunc(bz2name, "r:bz2") |
| testfunc(bz2name, "r|*") |
| testfunc(bz2name, "r|bz2") |
| |
| def test_detect_file(self): |
| self._test_modes(self._testfunc_file) |
| |
| def test_detect_fileobj(self): |
| self._test_modes(self._testfunc_fileobj) |
| |
| |
| class MemberReadTest(ReadTest): |
| |
| def _test_member(self, tarinfo, chksum=None, **kwargs): |
| if chksum is not None: |
| self.assertTrue(md5sum(self.tar.extractfile(tarinfo).read()) == chksum, |
| "wrong md5sum for %s" % tarinfo.name) |
| |
| kwargs["mtime"] = 0o7606136617 |
| kwargs["uid"] = 1000 |
| kwargs["gid"] = 100 |
| if "old-v7" not in tarinfo.name: |
| # V7 tar can't handle alphabetic owners. |
| kwargs["uname"] = "tarfile" |
| kwargs["gname"] = "tarfile" |
| for k, v in kwargs.items(): |
| self.assertTrue(getattr(tarinfo, k) == v, |
| "wrong value in %s field of %s" % (k, tarinfo.name)) |
| |
| def test_find_regtype(self): |
| tarinfo = self.tar.getmember("ustar/regtype") |
| self._test_member(tarinfo, size=7011, chksum=md5_regtype) |
| |
| def test_find_conttype(self): |
| tarinfo = self.tar.getmember("ustar/conttype") |
| self._test_member(tarinfo, size=7011, chksum=md5_regtype) |
| |
| def test_find_dirtype(self): |
| tarinfo = self.tar.getmember("ustar/dirtype") |
| self._test_member(tarinfo, size=0) |
| |
| def test_find_dirtype_with_size(self): |
| tarinfo = self.tar.getmember("ustar/dirtype-with-size") |
| self._test_member(tarinfo, size=255) |
| |
| def test_find_lnktype(self): |
| tarinfo = self.tar.getmember("ustar/lnktype") |
| self._test_member(tarinfo, size=0, linkname="ustar/regtype") |
| |
| def test_find_symtype(self): |
| tarinfo = self.tar.getmember("ustar/symtype") |
| self._test_member(tarinfo, size=0, linkname="regtype") |
| |
| def test_find_blktype(self): |
| tarinfo = self.tar.getmember("ustar/blktype") |
| self._test_member(tarinfo, size=0, devmajor=3, devminor=0) |
| |
| def test_find_chrtype(self): |
| tarinfo = self.tar.getmember("ustar/chrtype") |
| self._test_member(tarinfo, size=0, devmajor=1, devminor=3) |
| |
| def test_find_fifotype(self): |
| tarinfo = self.tar.getmember("ustar/fifotype") |
| self._test_member(tarinfo, size=0) |
| |
| def test_find_sparse(self): |
| tarinfo = self.tar.getmember("ustar/sparse") |
| self._test_member(tarinfo, size=86016, chksum=md5_sparse) |
| |
| def test_find_umlauts(self): |
| tarinfo = self.tar.getmember("ustar/umlauts-\xc4\xd6\xdc\xe4\xf6\xfc\xdf") |
| self._test_member(tarinfo, size=7011, chksum=md5_regtype) |
| |
| def test_find_ustar_longname(self): |
| name = "ustar/" + "12345/" * 39 + "1234567/longname" |
| self.assertIn(name, self.tar.getnames()) |
| |
| def test_find_regtype_oldv7(self): |
| tarinfo = self.tar.getmember("misc/regtype-old-v7") |
| self._test_member(tarinfo, size=7011, chksum=md5_regtype) |
| |
| def test_find_pax_umlauts(self): |
| self.tar = tarfile.open(self.tarname, mode=self.mode, encoding="iso8859-1") |
| tarinfo = self.tar.getmember("pax/umlauts-\xc4\xd6\xdc\xe4\xf6\xfc\xdf") |
| self._test_member(tarinfo, size=7011, chksum=md5_regtype) |
| |
| |
| class LongnameTest(ReadTest): |
| |
| def test_read_longname(self): |
| # Test reading of longname (bug #1471427). |
| longname = self.subdir + "/" + "123/" * 125 + "longname" |
| try: |
| tarinfo = self.tar.getmember(longname) |
| except KeyError: |
| self.fail("longname not found") |
| self.assertTrue(tarinfo.type != tarfile.DIRTYPE, "read longname as dirtype") |
| |
| def test_read_longlink(self): |
| longname = self.subdir + "/" + "123/" * 125 + "longname" |
| longlink = self.subdir + "/" + "123/" * 125 + "longlink" |
| try: |
| tarinfo = self.tar.getmember(longlink) |
| except KeyError: |
| self.fail("longlink not found") |
| self.assertTrue(tarinfo.linkname == longname, "linkname wrong") |
| |
| def test_truncated_longname(self): |
| longname = self.subdir + "/" + "123/" * 125 + "longname" |
| tarinfo = self.tar.getmember(longname) |
| offset = tarinfo.offset |
| self.tar.fileobj.seek(offset) |
| fobj = io.BytesIO(self.tar.fileobj.read(3 * 512)) |
| self.assertRaises(tarfile.ReadError, tarfile.open, name="foo.tar", fileobj=fobj) |
| |
| def test_header_offset(self): |
| # Test if the start offset of the TarInfo object includes |
| # the preceding extended header. |
| longname = self.subdir + "/" + "123/" * 125 + "longname" |
| offset = self.tar.getmember(longname).offset |
| fobj = open(tarname, "rb") |
| fobj.seek(offset) |
| tarinfo = tarfile.TarInfo.frombuf(fobj.read(512), "iso8859-1", "strict") |
| self.assertEqual(tarinfo.type, self.longnametype) |
| |
| |
| class GNUReadTest(LongnameTest): |
| |
| subdir = "gnu" |
| longnametype = tarfile.GNUTYPE_LONGNAME |
| |
| def test_sparse_file(self): |
| tarinfo1 = self.tar.getmember("ustar/sparse") |
| fobj1 = self.tar.extractfile(tarinfo1) |
| tarinfo2 = self.tar.getmember("gnu/sparse") |
| fobj2 = self.tar.extractfile(tarinfo2) |
| self.assertEqual(fobj1.read(), fobj2.read(), |
| "sparse file extraction failed") |
| |
| |
| class PaxReadTest(LongnameTest): |
| |
| subdir = "pax" |
| longnametype = tarfile.XHDTYPE |
| |
| def test_pax_global_headers(self): |
| tar = tarfile.open(tarname, encoding="iso8859-1") |
| |
| tarinfo = tar.getmember("pax/regtype1") |
| self.assertEqual(tarinfo.uname, "foo") |
| self.assertEqual(tarinfo.gname, "bar") |
| self.assertEqual(tarinfo.pax_headers.get("VENDOR.umlauts"), "\xc4\xd6\xdc\xe4\xf6\xfc\xdf") |
| |
| tarinfo = tar.getmember("pax/regtype2") |
| self.assertEqual(tarinfo.uname, "") |
| self.assertEqual(tarinfo.gname, "bar") |
| self.assertEqual(tarinfo.pax_headers.get("VENDOR.umlauts"), "\xc4\xd6\xdc\xe4\xf6\xfc\xdf") |
| |
| tarinfo = tar.getmember("pax/regtype3") |
| self.assertEqual(tarinfo.uname, "tarfile") |
| self.assertEqual(tarinfo.gname, "tarfile") |
| self.assertEqual(tarinfo.pax_headers.get("VENDOR.umlauts"), "\xc4\xd6\xdc\xe4\xf6\xfc\xdf") |
| |
| def test_pax_number_fields(self): |
| # All following number fields are read from the pax header. |
| tar = tarfile.open(tarname, encoding="iso8859-1") |
| tarinfo = tar.getmember("pax/regtype4") |
| self.assertEqual(tarinfo.size, 7011) |
| self.assertEqual(tarinfo.uid, 123) |
| self.assertEqual(tarinfo.gid, 123) |
| self.assertEqual(tarinfo.mtime, 1041808783.0) |
| self.assertEqual(type(tarinfo.mtime), float) |
| self.assertEqual(float(tarinfo.pax_headers["atime"]), 1041808783.0) |
| self.assertEqual(float(tarinfo.pax_headers["ctime"]), 1041808783.0) |
| |
| |
| class WriteTestBase(unittest.TestCase): |
| # Put all write tests in here that are supposed to be tested |
| # in all possible mode combinations. |
| |
| def test_fileobj_no_close(self): |
| fobj = io.BytesIO() |
| tar = tarfile.open(fileobj=fobj, mode=self.mode) |
| tar.addfile(tarfile.TarInfo("foo")) |
| tar.close() |
| self.assertTrue(fobj.closed is False, "external fileobjs must never closed") |
| |
| |
| class WriteTest(WriteTestBase): |
| |
| mode = "w:" |
| |
| def test_100_char_name(self): |
| # The name field in a tar header stores strings of at most 100 chars. |
| # If a string is shorter than 100 chars it has to be padded with '\0', |
| # which implies that a string of exactly 100 chars is stored without |
| # a trailing '\0'. |
| name = "0123456789" * 10 |
| tar = tarfile.open(tmpname, self.mode) |
| t = tarfile.TarInfo(name) |
| tar.addfile(t) |
| tar.close() |
| |
| tar = tarfile.open(tmpname) |
| self.assertTrue(tar.getnames()[0] == name, |
| "failed to store 100 char filename") |
| tar.close() |
| |
| def test_tar_size(self): |
| # Test for bug #1013882. |
| tar = tarfile.open(tmpname, self.mode) |
| path = os.path.join(TEMPDIR, "file") |
| fobj = open(path, "wb") |
| fobj.write(b"aaa") |
| fobj.close() |
| tar.add(path) |
| tar.close() |
| self.assertTrue(os.path.getsize(tmpname) > 0, |
| "tarfile is empty") |
| |
| # The test_*_size tests test for bug #1167128. |
| def test_file_size(self): |
| tar = tarfile.open(tmpname, self.mode) |
| |
| path = os.path.join(TEMPDIR, "file") |
| fobj = open(path, "wb") |
| fobj.close() |
| tarinfo = tar.gettarinfo(path) |
| self.assertEqual(tarinfo.size, 0) |
| |
| fobj = open(path, "wb") |
| fobj.write(b"aaa") |
| fobj.close() |
| tarinfo = tar.gettarinfo(path) |
| self.assertEqual(tarinfo.size, 3) |
| |
| tar.close() |
| |
| def test_directory_size(self): |
| path = os.path.join(TEMPDIR, "directory") |
| os.mkdir(path) |
| try: |
| tar = tarfile.open(tmpname, self.mode) |
| tarinfo = tar.gettarinfo(path) |
| self.assertEqual(tarinfo.size, 0) |
| finally: |
| os.rmdir(path) |
| |
| def test_link_size(self): |
| if hasattr(os, "link"): |
| link = os.path.join(TEMPDIR, "link") |
| target = os.path.join(TEMPDIR, "link_target") |
| open(target, "wb").close() |
| os.link(target, link) |
| try: |
| tar = tarfile.open(tmpname, self.mode) |
| tarinfo = tar.gettarinfo(link) |
| self.assertEqual(tarinfo.size, 0) |
| finally: |
| os.remove(target) |
| os.remove(link) |
| |
| def test_symlink_size(self): |
| if hasattr(os, "symlink"): |
| path = os.path.join(TEMPDIR, "symlink") |
| os.symlink("link_target", path) |
| try: |
| tar = tarfile.open(tmpname, self.mode) |
| tarinfo = tar.gettarinfo(path) |
| self.assertEqual(tarinfo.size, 0) |
| finally: |
| os.remove(path) |
| |
| def test_add_self(self): |
| # Test for #1257255. |
| dstname = os.path.abspath(tmpname) |
| |
| tar = tarfile.open(tmpname, self.mode) |
| self.assertTrue(tar.name == dstname, "archive name must be absolute") |
| |
| tar.add(dstname) |
| self.assertTrue(tar.getnames() == [], "added the archive to itself") |
| |
| cwd = os.getcwd() |
| os.chdir(TEMPDIR) |
| tar.add(dstname) |
| os.chdir(cwd) |
| self.assertTrue(tar.getnames() == [], "added the archive to itself") |
| |
| def test_exclude(self): |
| tempdir = os.path.join(TEMPDIR, "exclude") |
| os.mkdir(tempdir) |
| try: |
| for name in ("foo", "bar", "baz"): |
| name = os.path.join(tempdir, name) |
| open(name, "wb").close() |
| |
| exclude = os.path.isfile |
| |
| tar = tarfile.open(tmpname, self.mode, encoding="iso8859-1") |
| with support.check_warnings(("use the filter argument", |
| DeprecationWarning)): |
| tar.add(tempdir, arcname="empty_dir", exclude=exclude) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, "r") |
| self.assertEqual(len(tar.getmembers()), 1) |
| self.assertEqual(tar.getnames()[0], "empty_dir") |
| finally: |
| shutil.rmtree(tempdir) |
| |
| def test_filter(self): |
| tempdir = os.path.join(TEMPDIR, "filter") |
| os.mkdir(tempdir) |
| try: |
| for name in ("foo", "bar", "baz"): |
| name = os.path.join(tempdir, name) |
| open(name, "wb").close() |
| |
| def filter(tarinfo): |
| if os.path.basename(tarinfo.name) == "bar": |
| return |
| tarinfo.uid = 123 |
| tarinfo.uname = "foo" |
| return tarinfo |
| |
| tar = tarfile.open(tmpname, self.mode, encoding="iso8859-1") |
| tar.add(tempdir, arcname="empty_dir", filter=filter) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, "r") |
| for tarinfo in tar: |
| self.assertEqual(tarinfo.uid, 123) |
| self.assertEqual(tarinfo.uname, "foo") |
| self.assertEqual(len(tar.getmembers()), 3) |
| tar.close() |
| finally: |
| shutil.rmtree(tempdir) |
| |
| # Guarantee that stored pathnames are not modified. Don't |
| # remove ./ or ../ or double slashes. Still make absolute |
| # pathnames relative. |
| # For details see bug #6054. |
| def _test_pathname(self, path, cmp_path=None, dir=False): |
| # Create a tarfile with an empty member named path |
| # and compare the stored name with the original. |
| foo = os.path.join(TEMPDIR, "foo") |
| if not dir: |
| open(foo, "w").close() |
| else: |
| os.mkdir(foo) |
| |
| tar = tarfile.open(tmpname, self.mode) |
| tar.add(foo, arcname=path) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, "r") |
| t = tar.next() |
| tar.close() |
| |
| if not dir: |
| os.remove(foo) |
| else: |
| os.rmdir(foo) |
| |
| self.assertEqual(t.name, cmp_path or path.replace(os.sep, "/")) |
| |
| def test_pathnames(self): |
| self._test_pathname("foo") |
| self._test_pathname(os.path.join("foo", ".", "bar")) |
| self._test_pathname(os.path.join("foo", "..", "bar")) |
| self._test_pathname(os.path.join(".", "foo")) |
| self._test_pathname(os.path.join(".", "foo", ".")) |
| self._test_pathname(os.path.join(".", "foo", ".", "bar")) |
| self._test_pathname(os.path.join(".", "foo", "..", "bar")) |
| self._test_pathname(os.path.join(".", "foo", "..", "bar")) |
| self._test_pathname(os.path.join("..", "foo")) |
| self._test_pathname(os.path.join("..", "foo", "..")) |
| self._test_pathname(os.path.join("..", "foo", ".", "bar")) |
| self._test_pathname(os.path.join("..", "foo", "..", "bar")) |
| |
| self._test_pathname("foo" + os.sep + os.sep + "bar") |
| self._test_pathname("foo" + os.sep + os.sep, "foo", dir=True) |
| |
| def test_abs_pathnames(self): |
| if sys.platform == "win32": |
| self._test_pathname("C:\\foo", "foo") |
| else: |
| self._test_pathname("/foo", "foo") |
| self._test_pathname("///foo", "foo") |
| |
| def test_cwd(self): |
| # Test adding the current working directory. |
| cwd = os.getcwd() |
| os.chdir(TEMPDIR) |
| try: |
| open("foo", "w").close() |
| |
| tar = tarfile.open(tmpname, self.mode) |
| tar.add(".") |
| tar.close() |
| |
| tar = tarfile.open(tmpname, "r") |
| for t in tar: |
| self.assert_(t.name == "." or t.name.startswith("./")) |
| tar.close() |
| finally: |
| os.chdir(cwd) |
| |
| |
| class StreamWriteTest(WriteTestBase): |
| |
| mode = "w|" |
| |
| def test_stream_padding(self): |
| # Test for bug #1543303. |
| tar = tarfile.open(tmpname, self.mode) |
| tar.close() |
| |
| if self.mode.endswith("gz"): |
| fobj = gzip.GzipFile(tmpname) |
| data = fobj.read() |
| fobj.close() |
| elif self.mode.endswith("bz2"): |
| dec = bz2.BZ2Decompressor() |
| data = open(tmpname, "rb").read() |
| data = dec.decompress(data) |
| self.assertTrue(len(dec.unused_data) == 0, |
| "found trailing data") |
| else: |
| fobj = open(tmpname, "rb") |
| data = fobj.read() |
| fobj.close() |
| |
| self.assertTrue(data.count(b"\0") == tarfile.RECORDSIZE, |
| "incorrect zero padding") |
| |
| |
| class GNUWriteTest(unittest.TestCase): |
| # This testcase checks for correct creation of GNU Longname |
| # and Longlink extended headers (cp. bug #812325). |
| |
| def _length(self, s): |
| blocks, remainder = divmod(len(s) + 1, 512) |
| if remainder: |
| blocks += 1 |
| return blocks * 512 |
| |
| def _calc_size(self, name, link=None): |
| # Initial tar header |
| count = 512 |
| |
| if len(name) > tarfile.LENGTH_NAME: |
| # GNU longname extended header + longname |
| count += 512 |
| count += self._length(name) |
| if link is not None and len(link) > tarfile.LENGTH_LINK: |
| # GNU longlink extended header + longlink |
| count += 512 |
| count += self._length(link) |
| return count |
| |
| def _test(self, name, link=None): |
| tarinfo = tarfile.TarInfo(name) |
| if link: |
| tarinfo.linkname = link |
| tarinfo.type = tarfile.LNKTYPE |
| |
| tar = tarfile.open(tmpname, "w") |
| tar.format = tarfile.GNU_FORMAT |
| tar.addfile(tarinfo) |
| |
| v1 = self._calc_size(name, link) |
| v2 = tar.offset |
| self.assertTrue(v1 == v2, "GNU longname/longlink creation failed") |
| |
| tar.close() |
| |
| tar = tarfile.open(tmpname) |
| member = tar.next() |
| self.assertIsNotNone(member, |
| "unable to read longname member") |
| self.assertEqual(tarinfo.name, member.name, |
| "unable to read longname member") |
| self.assertEqual(tarinfo.linkname, member.linkname, |
| "unable to read longname member") |
| |
| def test_longname_1023(self): |
| self._test(("longnam/" * 127) + "longnam") |
| |
| def test_longname_1024(self): |
| self._test(("longnam/" * 127) + "longname") |
| |
| def test_longname_1025(self): |
| self._test(("longnam/" * 127) + "longname_") |
| |
| def test_longlink_1023(self): |
| self._test("name", ("longlnk/" * 127) + "longlnk") |
| |
| def test_longlink_1024(self): |
| self._test("name", ("longlnk/" * 127) + "longlink") |
| |
| def test_longlink_1025(self): |
| self._test("name", ("longlnk/" * 127) + "longlink_") |
| |
| def test_longnamelink_1023(self): |
| self._test(("longnam/" * 127) + "longnam", |
| ("longlnk/" * 127) + "longlnk") |
| |
| def test_longnamelink_1024(self): |
| self._test(("longnam/" * 127) + "longname", |
| ("longlnk/" * 127) + "longlink") |
| |
| def test_longnamelink_1025(self): |
| self._test(("longnam/" * 127) + "longname_", |
| ("longlnk/" * 127) + "longlink_") |
| |
| |
| class HardlinkTest(unittest.TestCase): |
| # Test the creation of LNKTYPE (hardlink) members in an archive. |
| |
| def setUp(self): |
| self.foo = os.path.join(TEMPDIR, "foo") |
| self.bar = os.path.join(TEMPDIR, "bar") |
| |
| fobj = open(self.foo, "wb") |
| fobj.write(b"foo") |
| fobj.close() |
| |
| os.link(self.foo, self.bar) |
| |
| self.tar = tarfile.open(tmpname, "w") |
| self.tar.add(self.foo) |
| |
| def tearDown(self): |
| self.tar.close() |
| os.remove(self.foo) |
| os.remove(self.bar) |
| |
| def test_add_twice(self): |
| # The same name will be added as a REGTYPE every |
| # time regardless of st_nlink. |
| tarinfo = self.tar.gettarinfo(self.foo) |
| self.assertTrue(tarinfo.type == tarfile.REGTYPE, |
| "add file as regular failed") |
| |
| def test_add_hardlink(self): |
| tarinfo = self.tar.gettarinfo(self.bar) |
| self.assertTrue(tarinfo.type == tarfile.LNKTYPE, |
| "add file as hardlink failed") |
| |
| def test_dereference_hardlink(self): |
| self.tar.dereference = True |
| tarinfo = self.tar.gettarinfo(self.bar) |
| self.assertTrue(tarinfo.type == tarfile.REGTYPE, |
| "dereferencing hardlink failed") |
| |
| |
| class PaxWriteTest(GNUWriteTest): |
| |
| def _test(self, name, link=None): |
| # See GNUWriteTest. |
| tarinfo = tarfile.TarInfo(name) |
| if link: |
| tarinfo.linkname = link |
| tarinfo.type = tarfile.LNKTYPE |
| |
| tar = tarfile.open(tmpname, "w", format=tarfile.PAX_FORMAT) |
| tar.addfile(tarinfo) |
| tar.close() |
| |
| tar = tarfile.open(tmpname) |
| if link: |
| l = tar.getmembers()[0].linkname |
| self.assertTrue(link == l, "PAX longlink creation failed") |
| else: |
| n = tar.getmembers()[0].name |
| self.assertTrue(name == n, "PAX longname creation failed") |
| |
| def test_pax_global_header(self): |
| pax_headers = { |
| "foo": "bar", |
| "uid": "0", |
| "mtime": "1.23", |
| "test": "\xe4\xf6\xfc", |
| "\xe4\xf6\xfc": "test"} |
| |
| tar = tarfile.open(tmpname, "w", format=tarfile.PAX_FORMAT, |
| pax_headers=pax_headers) |
| tar.addfile(tarfile.TarInfo("test")) |
| tar.close() |
| |
| # Test if the global header was written correctly. |
| tar = tarfile.open(tmpname, encoding="iso8859-1") |
| self.assertEqual(tar.pax_headers, pax_headers) |
| self.assertEqual(tar.getmembers()[0].pax_headers, pax_headers) |
| |
| # Test if all the fields are strings. |
| for key, val in tar.pax_headers.items(): |
| self.assertTrue(type(key) is not bytes) |
| self.assertTrue(type(val) is not bytes) |
| if key in tarfile.PAX_NUMBER_FIELDS: |
| try: |
| tarfile.PAX_NUMBER_FIELDS[key](val) |
| except (TypeError, ValueError): |
| self.fail("unable to convert pax header field") |
| |
| def test_pax_extended_header(self): |
| # The fields from the pax header have priority over the |
| # TarInfo. |
| pax_headers = {"path": "foo", "uid": "123"} |
| |
| tar = tarfile.open(tmpname, "w", format=tarfile.PAX_FORMAT, encoding="iso8859-1") |
| t = tarfile.TarInfo() |
| t.name = "\xe4\xf6\xfc" # non-ASCII |
| t.uid = 8**8 # too large |
| t.pax_headers = pax_headers |
| tar.addfile(t) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, encoding="iso8859-1") |
| t = tar.getmembers()[0] |
| self.assertEqual(t.pax_headers, pax_headers) |
| self.assertEqual(t.name, "foo") |
| self.assertEqual(t.uid, 123) |
| |
| |
| class UstarUnicodeTest(unittest.TestCase): |
| |
| format = tarfile.USTAR_FORMAT |
| |
| def test_iso8859_1_filename(self): |
| self._test_unicode_filename("iso8859-1") |
| |
| def test_utf7_filename(self): |
| self._test_unicode_filename("utf7") |
| |
| def test_utf8_filename(self): |
| self._test_unicode_filename("utf8") |
| |
| def _test_unicode_filename(self, encoding): |
| tar = tarfile.open(tmpname, "w", format=self.format, encoding=encoding, errors="strict") |
| name = "\xe4\xf6\xfc" |
| tar.addfile(tarfile.TarInfo(name)) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, encoding=encoding) |
| self.assertEqual(tar.getmembers()[0].name, name) |
| tar.close() |
| |
| def test_unicode_filename_error(self): |
| if self.format == tarfile.PAX_FORMAT: |
| # PAX_FORMAT ignores encoding in write mode. |
| return |
| |
| tar = tarfile.open(tmpname, "w", format=self.format, encoding="ascii", errors="strict") |
| tarinfo = tarfile.TarInfo() |
| |
| tarinfo.name = "\xe4\xf6\xfc" |
| self.assertRaises(UnicodeError, tar.addfile, tarinfo) |
| |
| tarinfo.name = "foo" |
| tarinfo.uname = "\xe4\xf6\xfc" |
| self.assertRaises(UnicodeError, tar.addfile, tarinfo) |
| |
| def test_unicode_argument(self): |
| tar = tarfile.open(tarname, "r", encoding="iso8859-1", errors="strict") |
| for t in tar: |
| self.assertTrue(type(t.name) is str) |
| self.assertTrue(type(t.linkname) is str) |
| self.assertTrue(type(t.uname) is str) |
| self.assertTrue(type(t.gname) is str) |
| tar.close() |
| |
| def test_uname_unicode(self): |
| t = tarfile.TarInfo("foo") |
| t.uname = "\xe4\xf6\xfc" |
| t.gname = "\xe4\xf6\xfc" |
| |
| tar = tarfile.open(tmpname, mode="w", format=self.format, encoding="iso8859-1") |
| tar.addfile(t) |
| tar.close() |
| |
| tar = tarfile.open(tmpname, encoding="iso8859-1") |
| t = tar.getmember("foo") |
| self.assertEqual(t.uname, "\xe4\xf6\xfc") |
| self.assertEqual(t.gname, "\xe4\xf6\xfc") |
| |
| if self.format != tarfile.PAX_FORMAT: |
| tar = tarfile.open(tmpname, encoding="ascii") |
| t = tar.getmember("foo") |
| self.assertEqual(t.uname, "\ufffd\ufffd\ufffd") |
| self.assertEqual(t.gname, "\ufffd\ufffd\ufffd") |
| |
| |
| class GNUUnicodeTest(UstarUnicodeTest): |
| |
| format = tarfile.GNU_FORMAT |
| |
| |
| class PAXUnicodeTest(UstarUnicodeTest): |
| |
| format = tarfile.PAX_FORMAT |
| |
| |
| class AppendTest(unittest.TestCase): |
| # Test append mode (cp. patch #1652681). |
| |
| def setUp(self): |
| self.tarname = tmpname |
| if os.path.exists(self.tarname): |
| os.remove(self.tarname) |
| |
| def _add_testfile(self, fileobj=None): |
| tar = tarfile.open(self.tarname, "a", fileobj=fileobj) |
| tar.addfile(tarfile.TarInfo("bar")) |
| tar.close() |
| |
| def _create_testtar(self, mode="w:"): |
| src = tarfile.open(tarname, encoding="iso8859-1") |
| t = src.getmember("ustar/regtype") |
| t.name = "foo" |
| f = src.extractfile(t) |
| tar = tarfile.open(self.tarname, mode) |
| tar.addfile(t, f) |
| tar.close() |
| |
| def _test(self, names=["bar"], fileobj=None): |
| tar = tarfile.open(self.tarname, fileobj=fileobj) |
| self.assertEqual(tar.getnames(), names) |
| |
| def test_non_existing(self): |
| self._add_testfile() |
| self._test() |
| |
| def test_empty(self): |
| tarfile.open(self.tarname, "w:").close() |
| self._add_testfile() |
| self._test() |
| |
| def test_empty_fileobj(self): |
| fobj = io.BytesIO(b"\0" * 1024) |
| self._add_testfile(fobj) |
| fobj.seek(0) |
| self._test(fileobj=fobj) |
| |
| def test_fileobj(self): |
| self._create_testtar() |
| data = open(self.tarname, "rb").read() |
| fobj = io.BytesIO(data) |
| self._add_testfile(fobj) |
| fobj.seek(0) |
| self._test(names=["foo", "bar"], fileobj=fobj) |
| |
| def test_existing(self): |
| self._create_testtar() |
| self._add_testfile() |
| self._test(names=["foo", "bar"]) |
| |
| def test_append_gz(self): |
| if gzip is None: |
| return |
| self._create_testtar("w:gz") |
| self.assertRaises(tarfile.ReadError, tarfile.open, tmpname, "a") |
| |
| def test_append_bz2(self): |
| if bz2 is None: |
| return |
| self._create_testtar("w:bz2") |
| self.assertRaises(tarfile.ReadError, tarfile.open, tmpname, "a") |
| |
| # Append mode is supposed to fail if the tarfile to append to |
| # does not end with a zero block. |
| def _test_error(self, data): |
| open(self.tarname, "wb").write(data) |
| self.assertRaises(tarfile.ReadError, self._add_testfile) |
| |
| def test_null(self): |
| self._test_error(b"") |
| |
| def test_incomplete(self): |
| self._test_error(b"\0" * 13) |
| |
| def test_premature_eof(self): |
| data = tarfile.TarInfo("foo").tobuf() |
| self._test_error(data) |
| |
| def test_trailing_garbage(self): |
| data = tarfile.TarInfo("foo").tobuf() |
| self._test_error(data + b"\0" * 13) |
| |
| def test_invalid(self): |
| self._test_error(b"a" * 512) |
| |
| |
| class LimitsTest(unittest.TestCase): |
| |
| def test_ustar_limits(self): |
| # 100 char name |
| tarinfo = tarfile.TarInfo("0123456789" * 10) |
| tarinfo.tobuf(tarfile.USTAR_FORMAT) |
| |
| # 101 char name that cannot be stored |
| tarinfo = tarfile.TarInfo("0123456789" * 10 + "0") |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.USTAR_FORMAT) |
| |
| # 256 char name with a slash at pos 156 |
| tarinfo = tarfile.TarInfo("123/" * 62 + "longname") |
| tarinfo.tobuf(tarfile.USTAR_FORMAT) |
| |
| # 256 char name that cannot be stored |
| tarinfo = tarfile.TarInfo("1234567/" * 31 + "longname") |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.USTAR_FORMAT) |
| |
| # 512 char name |
| tarinfo = tarfile.TarInfo("123/" * 126 + "longname") |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.USTAR_FORMAT) |
| |
| # 512 char linkname |
| tarinfo = tarfile.TarInfo("longlink") |
| tarinfo.linkname = "123/" * 126 + "longname" |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.USTAR_FORMAT) |
| |
| # uid > 8 digits |
| tarinfo = tarfile.TarInfo("name") |
| tarinfo.uid = 0o10000000 |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.USTAR_FORMAT) |
| |
| def test_gnu_limits(self): |
| tarinfo = tarfile.TarInfo("123/" * 126 + "longname") |
| tarinfo.tobuf(tarfile.GNU_FORMAT) |
| |
| tarinfo = tarfile.TarInfo("longlink") |
| tarinfo.linkname = "123/" * 126 + "longname" |
| tarinfo.tobuf(tarfile.GNU_FORMAT) |
| |
| # uid >= 256 ** 7 |
| tarinfo = tarfile.TarInfo("name") |
| tarinfo.uid = 0o4000000000000000000 |
| self.assertRaises(ValueError, tarinfo.tobuf, tarfile.GNU_FORMAT) |
| |
| def test_pax_limits(self): |
| tarinfo = tarfile.TarInfo("123/" * 126 + "longname") |
| tarinfo.tobuf(tarfile.PAX_FORMAT) |
| |
| tarinfo = tarfile.TarInfo("longlink") |
| tarinfo.linkname = "123/" * 126 + "longname" |
| tarinfo.tobuf(tarfile.PAX_FORMAT) |
| |
| tarinfo = tarfile.TarInfo("name") |
| tarinfo.uid = 0o4000000000000000000 |
| tarinfo.tobuf(tarfile.PAX_FORMAT) |
| |
| |
| class MiscTest(unittest.TestCase): |
| |
| def test_char_fields(self): |
| self.assertEqual(tarfile.stn("foo", 8, "ascii", "strict"), b"foo\0\0\0\0\0") |
| self.assertEqual(tarfile.stn("foobar", 3, "ascii", "strict"), b"foo") |
| self.assertEqual(tarfile.nts(b"foo\0\0\0\0\0", "ascii", "strict"), "foo") |
| self.assertEqual(tarfile.nts(b"foo\0bar\0", "ascii", "strict"), "foo") |
| |
| def test_number_fields(self): |
| self.assertEqual(tarfile.itn(1), b"0000001\x00") |
| self.assertEqual(tarfile.itn(0xffffffff), b"\x80\x00\x00\x00\xff\xff\xff\xff") |
| |
| |
| class ContextManagerTest(unittest.TestCase): |
| |
| def test_basic(self): |
| with tarfile.open(tarname) as tar: |
| self.assertFalse(tar.closed, "closed inside runtime context") |
| self.assertTrue(tar.closed, "context manager failed") |
| |
| def test_closed(self): |
| # The __enter__() method is supposed to raise IOError |
| # if the TarFile object is already closed. |
| tar = tarfile.open(tarname) |
| tar.close() |
| with self.assertRaises(IOError): |
| with tar: |
| pass |
| |
| def test_exception(self): |
| # Test if the IOError exception is passed through properly. |
| with self.assertRaises(Exception) as exc: |
| with tarfile.open(tarname) as tar: |
| raise IOError |
| self.assertIsInstance(exc.exception, IOError, |
| "wrong exception raised in context manager") |
| self.assertTrue(tar.closed, "context manager failed") |
| |
| def test_no_eof(self): |
| # __exit__() must not write end-of-archive blocks if an |
| # exception was raised. |
| try: |
| with tarfile.open(tmpname, "w") as tar: |
| raise Exception |
| except: |
| pass |
| self.assertEqual(os.path.getsize(tmpname), 0, |
| "context manager wrote an end-of-archive block") |
| self.assertTrue(tar.closed, "context manager failed") |
| |
| def test_eof(self): |
| # __exit__() must write end-of-archive blocks, i.e. call |
| # TarFile.close() if there was no error. |
| with tarfile.open(tmpname, "w"): |
| pass |
| self.assertNotEqual(os.path.getsize(tmpname), 0, |
| "context manager wrote no end-of-archive block") |
| |
| def test_fileobj(self): |
| # Test that __exit__() did not close the external file |
| # object. |
| fobj = open(tmpname, "wb") |
| try: |
| with tarfile.open(fileobj=fobj, mode="w") as tar: |
| raise Exception |
| except: |
| pass |
| self.assertFalse(fobj.closed, "external file object was closed") |
| self.assertTrue(tar.closed, "context manager failed") |
| fobj.close() |
| |
| |
| class GzipMiscReadTest(MiscReadTest): |
| tarname = gzipname |
| mode = "r:gz" |
| class GzipUstarReadTest(UstarReadTest): |
| tarname = gzipname |
| mode = "r:gz" |
| class GzipStreamReadTest(StreamReadTest): |
| tarname = gzipname |
| mode = "r|gz" |
| class GzipWriteTest(WriteTest): |
| mode = "w:gz" |
| class GzipStreamWriteTest(StreamWriteTest): |
| mode = "w|gz" |
| |
| |
| class Bz2MiscReadTest(MiscReadTest): |
| tarname = bz2name |
| mode = "r:bz2" |
| class Bz2UstarReadTest(UstarReadTest): |
| tarname = bz2name |
| mode = "r:bz2" |
| class Bz2StreamReadTest(StreamReadTest): |
| tarname = bz2name |
| mode = "r|bz2" |
| class Bz2WriteTest(WriteTest): |
| mode = "w:bz2" |
| class Bz2StreamWriteTest(StreamWriteTest): |
| mode = "w|bz2" |
| |
| class Bz2PartialReadTest(unittest.TestCase): |
| # Issue5068: The _BZ2Proxy.read() method loops forever |
| # on an empty or partial bzipped file. |
| |
| def _test_partial_input(self, mode): |
| class MyBytesIO(io.BytesIO): |
| hit_eof = False |
| def read(self, n): |
| if self.hit_eof: |
| raise AssertionError("infinite loop detected in tarfile.open()") |
| self.hit_eof = self.tell() == len(self.getvalue()) |
| return super(MyBytesIO, self).read(n) |
| def seek(self, *args): |
| self.hit_eof = False |
| return super(MyBytesIO, self).seek(*args) |
| |
| data = bz2.compress(tarfile.TarInfo("foo").tobuf()) |
| for x in range(len(data) + 1): |
| try: |
| tarfile.open(fileobj=MyBytesIO(data[:x]), mode=mode) |
| except tarfile.ReadError: |
| pass # we have no interest in ReadErrors |
| |
| def test_partial_input(self): |
| self._test_partial_input("r") |
| |
| def test_partial_input_bz2(self): |
| self._test_partial_input("r:bz2") |
| |
| |
| def test_main(): |
| os.makedirs(TEMPDIR) |
| |
| tests = [ |
| UstarReadTest, |
| MiscReadTest, |
| StreamReadTest, |
| DetectReadTest, |
| MemberReadTest, |
| GNUReadTest, |
| PaxReadTest, |
| WriteTest, |
| StreamWriteTest, |
| GNUWriteTest, |
| PaxWriteTest, |
| UstarUnicodeTest, |
| GNUUnicodeTest, |
| PAXUnicodeTest, |
| AppendTest, |
| LimitsTest, |
| MiscTest, |
| ContextManagerTest, |
| ] |
| |
| if hasattr(os, "link"): |
| tests.append(HardlinkTest) |
| |
| fobj = open(tarname, "rb") |
| data = fobj.read() |
| fobj.close() |
| |
| if gzip: |
| # Create testtar.tar.gz and add gzip-specific tests. |
| tar = gzip.open(gzipname, "wb") |
| tar.write(data) |
| tar.close() |
| |
| tests += [ |
| GzipMiscReadTest, |
| GzipUstarReadTest, |
| GzipStreamReadTest, |
| GzipWriteTest, |
| GzipStreamWriteTest, |
| ] |
| |
| if bz2: |
| # Create testtar.tar.bz2 and add bz2-specific tests. |
| tar = bz2.BZ2File(bz2name, "wb") |
| tar.write(data) |
| tar.close() |
| |
| tests += [ |
| Bz2MiscReadTest, |
| Bz2UstarReadTest, |
| Bz2StreamReadTest, |
| Bz2WriteTest, |
| Bz2StreamWriteTest, |
| Bz2PartialReadTest, |
| ] |
| |
| try: |
| support.run_unittest(*tests) |
| finally: |
| if os.path.exists(TEMPDIR): |
| shutil.rmtree(TEMPDIR) |
| |
| if __name__ == "__main__": |
| test_main() |