Merge branch 'main' into argparse-lazy

Author: savannahostrowski

Doc/library/logging.handlers.rst

@@ -631,7 +631,8 @@ supports sending logging messages to a remote or local Unix syslog.
    the form of a ``(host, port)`` tuple.  If *address* is not specified,
    ``('localhost', 514)`` is used.  The address is used to open a socket.  An
    alternative to providing a ``(host, port)`` tuple is providing an address as a
-   string, for example '/dev/log'. In this case, a Unix domain socket is used to
+   string or a :class:`bytes` object, for example '/dev/log'.
+   In this case, a Unix domain socket is used to
    send the message to the syslog. If *facility* is not specified,
    :const:`LOG_USER` is used. The type of socket opened depends on the
    *socktype* argument, which defaults to :const:`socket.SOCK_DGRAM` and thus
@@ -664,6 +665,9 @@ supports sending logging messages to a remote or local Unix syslog.
    .. versionchanged:: 3.14
       *timeout* was added.
 
+   .. versionchanged:: next
+      *address* can now be a :class:`bytes` object.
+
    .. method:: close()
 
       Closes the socket to the remote host.

Lib/argparse.py

@@ -775,7 +775,35 @@ def _iter_indented_subactions(self, action):
 
     def _split_lines(self, text, width):
         text = self._whitespace_matcher.sub(' ', text).strip()
-        return textwrap.wrap(text, width)
+        decolored = self._decolor(text)
+        if decolored == text:
+            return textwrap.wrap(text, width)
+
+        # gh-142035: colors inflate textwrap's length counts, so wrap
+        # the decolored text and re-apply colors per word; if textwrap
+        # split a word, keep the plain lines (colors can't be mapped).
+        plain = self._whitespace_matcher.sub(' ', decolored).strip()
+        if not plain:
+            # nothing visible to wrap (e.g. an empty interpolated value)
+            return [text]
+        plain_lines = textwrap.wrap(plain, width)
+        plain_words = plain.split()
+        colored_words = text.split()
+        # Drop escape-only tokens (e.g. an empty interpolated value).
+        if len(colored_words) != len(plain_words):
+            colored_words = [
+                word for word in colored_words if self._decolor(word)
+            ]
+        colored_lines = []
+        start = 0
+        for plain_line in plain_lines:
+            plain_line_words = plain_line.split()
+            end = start + len(plain_line_words)
+            if plain_words[start:end] != plain_line_words:
+                return plain_lines
+            colored_lines.append(' '.join(colored_words[start:end]))
+            start = end
+        return colored_lines
 
     def _fill_text(self, text, width, indent):
         text = self._whitespace_matcher.sub(' ', text).strip()

Lib/logging/handlers.py

@@ -886,8 +886,9 @@ def __init__(self, address=('localhost', SYSLOG_UDP_PORT),
         """
         Initialize a handler.
 
-        If address is specified as a string, a UNIX socket is used. To log to a
-        local syslogd, "SysLogHandler(address="/dev/log")" can be used.
+        If address is specified as a string or bytes, a UNIX socket is used.
+        To log to a local syslogd, "SysLogHandler(address="/dev/log")" can be
+        used.
         If facility is not specified, LOG_USER is used. If socktype is
         specified as socket.SOCK_DGRAM or socket.SOCK_STREAM, that specific
         socket type will be used. For Unix sockets, you can also specify a
@@ -938,7 +939,7 @@ def createSocket(self):
         address = self.address
         socktype = self.socktype
 
-        if isinstance(address, str):
+        if not isinstance(address, (list, tuple)):
             self.unixsocket = True
             # Syslog server may be unavailable during handler initialisation.
             # C's openlog() function also ignores connection errors.

Lib/test/test_argparse.py

@@ -7585,6 +7585,62 @@ def test_argparse_color_custom_usage(self):
             ),
         )
 
+    def test_argparse_color_wrapping_matches_uncolored(self):
+        # gh-142035: color codes must not affect where help text wraps.
+        # Stripping the escapes from colored help must yield exactly the
+        # same text as the uncolored help across representative widths.
+        def build(color, path="output.txt"):
+            parser = argparse.ArgumentParser(prog="PROG", color=color)
+            parser.add_argument(
+                "--mode",
+                default="auto",
+                choices=("auto", "fast", "slow"),
+                help="select the operating mode from the available choices "
+                     "%(choices)s and note the default is %(default)s here",
+            )
+            parser.add_argument(
+                "--path",
+                default=path,
+                help="write output to %(default)s and continue processing",
+            )
+            return parser
+
+        env = self.enterContext(os_helper.EnvironmentVarGuard())
+        paths = (
+            "output.txt",
+            "/var/lib/application/cache/unusually_long_generated_filename",
+            "production-read-only-replica",
+        )
+        for path in paths:
+            for columns in ("80", "60", "45", "30", "20"):
+                with self.subTest(path=path, columns=columns):
+                    env["COLUMNS"] = columns
+                    colored = build(color=True, path=path).format_help()
+                    plain = build(color=False, path=path).format_help()
+                    self.assertIn(
+                        f"{self.theme.interpolated_value}auto"
+                        f"{self.theme.reset}",
+                        colored,
+                    )
+                    self.assertEqual(_colorize.decolor(colored), plain)
+
+    def test_argparse_color_preserved_when_wrapping_between_words(self):
+        parser = argparse.ArgumentParser(prog="PROG", color=True)
+        parser.add_argument(
+            "--mode", default="auto",
+            help="select the %(default)s operating mode from the available "
+                 "options and continue with several more words",
+        )
+
+        env = self.enterContext(os_helper.EnvironmentVarGuard())
+        env["COLUMNS"] = "40"
+        help_text = parser.format_help()
+
+        self.assertIn(
+            f"{self.theme.interpolated_value}auto{self.theme.reset}",
+            help_text,
+        )
+
     def test_custom_formatter_function(self):
         def custom_formatter(prog):
             return argparse.RawTextHelpFormatter(prog, indent_increment=5)

Lib/test/test_bytes.py

@@ -2226,6 +2226,46 @@ def __len__(self):
 
         self.assertRaises(BufferError, ba.hex, S(b':'))
 
+    def test_no_init_called(self):
+        # A bytearray created without calling bytearray.__init__
+        # should not crash the interpreter (see gh-153419).
+        def bytearray_new():
+            return bytearray.__new__(bytearray)
+
+        bytearray_new().insert(0, 1)
+        bytearray_new().extend(b"x")
+        bytearray_new().extend([1, 2, 3])
+        bytearray_new().resize(4)
+        bytearray_new().__init__(5)
+        bytearray_new().__init__(b"xyz")
+        bytearray_new().take_bytes()
+        bytearray_new().take_bytes(0)
+
+        a = bytearray_new()
+        a.append(1)
+
+        a = bytearray_new()
+        a += b"x"
+
+        a = bytearray_new()
+        a[:] = b"xyz"
+
+    def test_reinit_length(self):
+        # There is a shortcut taken when resizing, where alloc/2 < newsize.
+        # In this case, the existing buffer is reused, rather than reset.
+        # If this happens when newsize == 0 and alloc == 1, then various
+        # code assumptions can be violated.  This test should catch those
+        # in debug builds. (see gh-153419)
+        a = bytearray(1)
+        a.__init__()
+        self.assertEqual(a, b"")
+
+    def test_reinit_with_view(self):
+        a = bytearray()
+        with memoryview(a):
+            self.assertRaises(BufferError, a.__init__, "x", "ascii")
+        self.assertEqual(a, b"")
+
 
 class AssortedBytesTest(unittest.TestCase):
     #

Lib/test/test_gdb/test_pretty_print.py

@@ -114,29 +114,28 @@ def test_bytes(self):
     @support.requires_resource('cpu')
     def test_strings(self):
         'Verify the pretty-printing of unicode strings'
-        # We cannot simply call locale.getpreferredencoding() here,
-        # as GDB might have been linked against a different version
-        # of Python with a different encoding and coercion policy
-        # with respect to PEP 538 and PEP 540.
+        # gdb emits its output in the host charset, which is not necessarily the
+        # getpreferredencoding() of the (possibly differently coerced) embedded
+        # Python.
         stdout, stderr = run_gdb(
             '--eval-command',
-            'python import locale; print(locale.getpreferredencoding())')
+            'python import gdb; print(gdb.host_charset())')
 
-        encoding = stdout
+        encoding = stdout.strip()
         if stderr or not encoding:
             raise RuntimeError(
-                f'unable to determine the Python locale preferred encoding '
-                f'of embedded Python in GDB\n'
+                f'unable to determine the host charset of gdb\n'
                 f'stdout={stdout!r}\n'
                 f'stderr={stderr!r}')
 
         def check_repr(text):
             try:
                 text.encode(encoding)
-            except UnicodeEncodeError:
+            # LookupError or ValueError if the host charset is unknown or invalid.
+            except (UnicodeEncodeError, LookupError, ValueError):
                 self.assertGdbRepr(text, ascii(text))
             else:
-                self.assertGdbRepr(text)
+                self.assertGdbRepr(text, repr(text).encode(encoding).decode('ascii', 'surrogateescape'))
 
         self.assertGdbRepr('')
         self.assertGdbRepr('And now for something hopefully the same')

Lib/test/test_gdb/util.py

@@ -78,7 +78,7 @@ def run_gdb(*args, exitcode=0, check=True, **env_vars):
         stdin=subprocess.PIPE,
         stdout=subprocess.PIPE,
         stderr=subprocess.PIPE,
-        encoding="utf8", errors="backslashreplace",
+        encoding="ascii", errors="surrogateescape",
         env=env)
 
     stdout = proc.stdout

Lib/test/test_logging.py

@@ -2138,6 +2138,45 @@ def setUp(self):
         self.addCleanup(os_helper.unlink, self.address)
         SysLogHandlerTest.setUp(self)
 
+    def test_bytes_address(self):
+        # The Unix socket address can also be specified as bytes.
+        if self.server_exception:
+            self.skipTest(self.server_exception)
+        hdlr = logging.handlers.SysLogHandler(os.fsencode(self.address))
+        self.addCleanup(hdlr.close)
+        self.assertTrue(hdlr.unixsocket)
+        logger = logging.getLogger("slh-bytes")
+        logger.addHandler(hdlr)
+        self.addCleanup(logger.removeHandler, hdlr)
+        logger.error("sp\xe4m")
+        self.handled.wait(support.LONG_TIMEOUT)
+        self.assertEqual(self.log_output, b'<11>sp\xc3\xa4m\x00')
+
+@unittest.skipUnless(sys.platform in ('linux', 'android'),
+                     'Linux specific test')
+class AbstractNamespaceSysLogHandlerTest(BaseTest):
+
+    """Test for SysLogHandler with a socket in the abstract namespace."""
+
+    def check(self, address):
+        sock = socket.socket(socket.AF_UNIX, socket.SOCK_DGRAM)
+        self.addCleanup(sock.close)
+        sock.bind(address)
+        sock.settimeout(support.LONG_TIMEOUT)
+        hdlr = logging.handlers.SysLogHandler(address)
+        self.addCleanup(hdlr.close)
+        self.assertTrue(hdlr.unixsocket)
+        hdlr.emit(logging.makeLogRecord({'msg': 'sp\xe4m'}))
+        self.assertEqual(sock.recv(1024), b'<12>sp\xc3\xa4m\x00')
+
+    def test_str_address(self):
+        # A str address is encoded with the filesystem encoding.
+        self.check('\0' + os_helper.TESTFN)
+
+    def test_bytes_address(self):
+        # The name is an arbitrary byte sequence, it need not be decodable.
+        self.check(b'\0test_logging_%d_\xff\xfe' % os.getpid())
+
 @unittest.skipUnless(socket_helper.IPV6_ENABLED,
                      'IPv6 support required for this test.')
 class IPv6SysLogHandlerTest(SysLogHandlerTest):

Misc/NEWS.d/next/Core_and_Builtins/2026-07-11-01-17-57.gh-issue-153419.U8HJOJ.rst

@@ -0,0 +1 @@
+Fix multiple :class:`bytearray` crashes and reference leaks caused by skipping :meth:`~object.__init__` and broken state setup code.

Misc/NEWS.d/next/Library/2026-07-22-17-35-24.gh-issue-70990.xVHKwt.rst

@@ -0,0 +1,4 @@
+:class:`logging.handlers.SysLogHandler` now accepts a :class:`bytes` address
+of a Unix domain socket, including an address in the abstract namespace.
+Previously only :class:`str` was recognized,
+and a :class:`bytes` address raised :exc:`ValueError`.