Sync pre-release CPython 3.15 branch from GitHub (2026-08-03)

Summary:
Imported python/cpython `3.15.0b4+dev` from upstream rev [`5bc7459`](https://www.github.com/python/cpython/commit/5bc7459474b7d3391b1d566f83728d8822654d1a) (committed 2026-08-03 21:21:11+00:00).

# Commit Info

- Base: (`3.15.0b4+dev`) - [`07e73c0`](https://www.github.com/python/cpython/commit/07e73c01c5ba2f8ff6b5f237a8a87dfb88787383) (commit date: 2026-08-02 18:49:20+00:00)
- Imported: (`3.15.0b4+dev`) - [`5bc7459`](https://www.github.com/python/cpython/commit/5bc7459474b7d3391b1d566f83728d8822654d1a) (commit date: 2026-08-03 21:21:11+00:00)

# Noteworthy file changes

- Stdlib files (1 added, 1 removed)
- Low-signal files (7 added) (NEWS.d, docs, .github)

Complete list of added/removed files: https://www.internalfb.com/intern/everpaste/?color=0&handle=GBay8yVwo4cfQaEDALspVNZF0AUCbr0LAAAz

Differential Revision: D114690383

fbshipit-source-id: 9765de93257f2638e4a43d6bdb20b5c38aa962aa

Author: generatedunixname1734921407115435

Doc/library/asyncio-task.rst

@@ -433,6 +433,10 @@ unless it is :exc:`asyncio.CancelledError`,
 is also included in the exception group.
 The same special case is made for
 :exc:`KeyboardInterrupt` and :exc:`SystemExit` as in the previous paragraph.
+There is an additional special case made only for the body of the
+``async with``: if it raises :exc:`GeneratorExit` and none of the
+other tasks raise exceptions that would be reported, then the
+:exc:`GeneratorExit` is reraised.
 
 Task groups are careful not to mix up the internal cancellation used to
 "wake up" their :meth:`~object.__aexit__` with cancellation requests
@@ -456,6 +460,10 @@ reported by :meth:`asyncio.Task.cancelling`.
    Improved handling of simultaneous internal and external cancellations
    and correct preservation of cancellation counts.
 
+.. versionchanged:: 3.15
+
+   Addition of the special case for :exc:`GeneratorExit`.
+
 Sleeping
 ========
 

Lib/asyncio/taskgroups.py

@@ -174,10 +174,23 @@ async def _aexit(self, et, exc):
                 self._parent_task.uncancel()
                 self._parent_task.cancel()
             try:
-                raise BaseExceptionGroup(
-                    'unhandled errors in a TaskGroup',
-                    self._errors,
-                ) from None
+                # If the *only* error is a GeneratorExit from the body
+                # of the group, then instead of raising an
+                # ExceptionGroup we raise GeneratorExit. This ensures
+                # that async generators that use TaskGroup properly
+                # swallow the exception on `aclose()` while ensuring
+                # that no exceptions from subtasks are swallowed.
+                if (
+                    et is not None
+                    and issubclass(et, GeneratorExit)
+                    and len(self._errors) == 1
+                ):
+                    raise exc
+                else:
+                    raise BaseExceptionGroup(
+                        'unhandled errors in a TaskGroup',
+                        self._errors,
+                    ) from None
             finally:
                 exc = None
 

Lib/csv.py

@@ -247,6 +247,8 @@ def sniff(self, sample, delimiters=None):
         that order, no matter how many times each of them occurs.
         """
 
+        sample = sample.replace('\r\n', '\n').replace('\r', '\n')
+
         quotechar, doublequote, delimiter, skipinitialspace = \
                    self._guess_quote_and_delimiter(sample, delimiters)
         if not delimiter:
@@ -284,12 +286,16 @@ def _guess_quote_and_delimiter(self, data, delimiters):
         """
         import re
 
+        # The body of a quoted field ends at the first quote which is
+        # not doubled, as it does for a reader.  A lazy ".*?" scans to
+        # the end of the sample instead, from every start: quadratically.
+        body = r'(?:(?P=quote){2}|(?!(?P=quote)).)*+'
         matches = []
-        for restr in (r'(?P<delim>[^\w\n"\'])(?P<space> ?)(?P<quote>["\']).*?(?P=quote)(?P=delim)',   # ,".*?",
-                      r'(?:^|\n)(?P<quote>["\']).*?(?P=quote)(?P<delim>[^\w\n"\'])(?P<space> ?)',     #  ".*?",
-                      r'(?P<delim>[^\w\n"\'])(?P<space> ?)(?P<quote>["\']).*?(?P=quote)(?:$|\r|\n)',  # ,".*?"
-                      r'(?:^|\n)(?P<quote>["\']).*?(?P=quote)(?:$|\r|\n)'):                           #  ".*?" (no delim, no space)
-            regexp = re.compile(restr, re.DOTALL | re.MULTILINE)
+        for restr in (r'(?P<delim>[^\w\n"\'])(?P<space> ?)(?P<quote>["\'])%s(?P=quote)(?P=delim)',   # ,"...",
+                      r'(?:^|\n)(?P<quote>["\'])%s(?P=quote)(?P<delim>[^\w\n"\'])(?P<space> ?)',     #  "...",
+                      r'(?P<delim>[^\w\n"\'])(?P<space> ?)(?P<quote>["\'])%s(?P=quote)(?:$|\n)',  # ,"..."
+                      r'(?:^|\n)(?P<quote>["\'])%s(?P=quote)(?:$|\n)'):                           #  "..." (no delim, no space)
+            regexp = re.compile(restr % body, re.DOTALL | re.MULTILINE)
             matches = regexp.findall(data)
             if matches:
                 break
@@ -332,18 +338,22 @@ def _guess_quote_and_delimiter(self, data, delimiters):
             delim = ''
             skipinitialspace = 0
 
-        # if we see an extra quote between delimiters, we've got a
-        # double quoted format
-        dq_regexp = re.compile(
-                               r"((%(delim)s)|^)\W*%(quote)s[^%(delim)s\n]*%(quote)s[^%(delim)s\n]*%(quote)s\W*((%(delim)s)|$)" % \
-                               {'delim':re.escape(delim), 'quote':quotechar}, re.MULTILINE)
-
-
-
-        if dq_regexp.search(data):
-            doublequote = True
-        else:
-            doublequote = False
+        # A doubled quote character inside a quoted field means
+        # a double quoted format.  Match whole fields, so that a match
+        # cannot slide across field boundaries.
+        doublequote = False
+        if delim:
+            dq_regexp = re.compile(
+                    r"(?:(?<=%(delim)s)|^)%(space)s%(quote)s"     # ,"
+                    r"((?:%(quote)s%(quote)s|[^%(quote)s]++)*+)"  # the body
+                    r"%(quote)s(?:%(delim)s|$)"                   # ",
+                    % {'delim': re.escape(delim), 'quote': quotechar,
+                       # Skipping spaces after a space rescans them.
+                       'space': ' *+' if delim != ' ' else ''},
+                    re.MULTILINE)
+            dquotechar = quotechar * 2
+            doublequote = any(dquotechar in m[1]
+                              for m in dq_regexp.finditer(data))
 
         return (quotechar, doublequote, delim, skipinitialspace)
 

Lib/ensurepip/__init__.py

@@ -10,7 +10,7 @@
 
 
 __all__ = ["version", "bootstrap"]
-_PIP_VERSION = "26.1.2"
+_PIP_VERSION = "26.2"
 
 # Directory of system wheel packages. Some Linux distribution packaging
 # policies recommend against bundling dependencies. For example, Fedora

…pip/_bundled/pip-26.1.2-py3-none-any.whl …repip/_bundled/pip-26.2-py3-none-any.whlLib/ensurepip/_bundled/pip-26.1.2-py3-none-any.whl renamed to Lib/ensurepip/_bundled/pip-26.2-py3-none-any.whl Lib/ensurepip/_bundled/pip-26.1.2-py3-none-any.whl renamed to Lib/ensurepip/_bundled/pip-26.2-py3-none-any.whl

@@ -1227,6 +1227,72 @@ async def fn_3():
         self.assertEqual(await race(fn_1, fn_2, fn_3), 1)
         self.assertListEqual(record, ["1 started", "2 started", "3 started", "1 finished"])
 
+    async def test_taskgroup_generator_exit_01(self):
+        # GeneratorExit in a TaskGroup should be fine
+        async def gen():
+            yield 1
+
+        async def fn():
+            async with asyncio.TaskGroup() as tg:
+                async for n in gen():
+                    yield n
+
+        g = fn()
+        await g.asend(None)
+        await g.aclose()
+
+    async def test_taskgroup_generator_exit_02(self):
+        # A lone GeneratorExit in a task should still give an ExceptionGroup
+        async def t():
+            raise GeneratorExit
+
+        async def fn():
+            async with asyncio.TaskGroup() as tg:
+                tg.create_task(t())
+
+        with self.assertRaises(BaseExceptionGroup) as cm:
+            await fn()
+        self.assertEqual(get_error_types(cm.exception), {GeneratorExit})
+
+    async def test_taskgroup_generator_exit_03(self):
+        # A GeneratorExit in one task and an error in another should
+        # still give an ExceptionGroup
+        async def t1():
+            raise GeneratorExit
+
+        async def t2():
+            raise AssertionError('t2 failed')
+
+        async def fn():
+            async with asyncio.TaskGroup() as tg:
+                tg.create_task(t1())
+                tg.create_task(t2())
+
+        with self.assertRaises(BaseExceptionGroup) as cm:
+            await fn()
+
+        self.assertEqual(get_error_types(cm.exception), {GeneratorExit, AssertionError})
+
+    async def test_taskgroup_generator_exit_04(self):
+        event = asyncio.Event()
+        async def t():
+            event.set()
+            raise AssertionError('t failed')
+
+        async def fn():
+            async with asyncio.TaskGroup() as tg:
+                tg.create_task(t())
+                yield 1
+
+        g = fn()
+        await g.asend(None)
+        await event.wait()  # wait for t() to run
+
+        with self.assertRaises(BaseExceptionGroup) as cm:
+            await g.aclose()
+
+        self.assertEqual(get_error_types(cm.exception), {GeneratorExit, AssertionError})
+
 
 class TestTaskGroup(BaseTestTaskGroup, unittest.IsolatedAsyncioTestCase):
     loop_factory = asyncio.EventLoop

Lib/test/test_asyncio/test_taskgroups.py

@@ -1564,6 +1564,57 @@ def test_zero_mode_tie_order_colon_first(self):
             sniffer.sniff(sample)
 
 
+    def test_sniff_regex_backtracking(self):
+        # gh-109638: this artificial sample used to take minutes.
+        sniffer = csv.Sniffer()
+        sample = '"",' * 100 + '"' * 100 + '0' + '"' * 100 + '0'
+        self.assertEqual(sniffer.sniff(sample).delimiter, ',')
+
+    def test_sniff_doublequote_across_fields(self):
+        # A quoted field which contains the delimiter, followed by
+        # an empty quoted field, is not a doubled quote.
+        sniffer = csv.Sniffer()
+        sample = '",","",","\n' * 4
+        dialect = sniffer.sniff(sample)
+        self.assertEqual(dialect.delimiter, ',')
+        self.assertEqual(dialect.quotechar, '"')
+        self.assertIs(dialect.doublequote, False)
+        self.assertEqual(next(csv.reader(StringIO(sample), dialect)),
+                         [',', '', ','])
+
+    def test_sniff_doublequote_record_separators(self):
+        # The record separator ends a field as a delimiter does.
+        sniffer = csv.Sniffer()
+        for sep in '\n', '\r\n', '\r':
+            with self.subTest(sep=sep):
+                sample = ('x,"a""b"' + sep + 'y,"c"' + sep) * 2
+                self.assertIs(sniffer.sniff(sample).doublequote, True)
+                sample = ('"",","' + sep) * 4
+                self.assertIs(sniffer.sniff(sample).doublequote, False)
+
+    def test_sniff_single_column(self):
+        # This sample used to be quadratic.
+        sniffer = csv.Sniffer()
+        sample = '"a"\n' + ' ' * 100000
+        with self.assertRaisesRegex(csv.Error, "Could not determine delimiter"):
+            sniffer.sniff(sample, delimiters=',;')
+
+    def test_sniff_space_delimiter(self):
+        # This sample used to be quadratic.
+        sniffer = csv.Sniffer()
+        sample = '"a" "b"\n' + ' ' * 100000
+        dialect = sniffer.sniff(sample)
+        self.assertEqual(dialect.delimiter, ' ')
+        self.assertIs(dialect.doublequote, False)
+
+    def test_sniff_quoted_single_column(self):
+        # gh-98820: this sample used to take minutes.
+        sniffer = csv.Sniffer()
+        sample = '"abcdefghijklmnopqrstuvwxyz"\n' * 10000
+        with self.assertRaisesRegex(csv.Error, "Could not determine delimiter"):
+            sniffer.sniff(sample, delimiters=',:|\t')
+
+
 class NUL:
     def write(s, *args):
         pass

Lib/test/test_csv.py

@@ -996,6 +996,13 @@ def test_parent_parser_outlives_its_subparsers__chain(self):
         del parser
         del subparser
 
+    def test_subparser_inherits_reparse_deferral(self):
+        for enabled in (True, False):
+            parser = expat.ParserCreate()
+            parser.SetReparseDeferralEnabled(enabled)
+            subparser = parser.ExternalEntityParserCreate(None)
+            self.assertEqual(subparser.GetReparseDeferralEnabled(), enabled)
+
 
 class ExternalEntityParserCreateErrorTest(unittest.TestCase):
     """ExternalEntityParserCreate error paths should not crash or leak

Lib/test/test_pyexpat.py

@@ -10,6 +10,7 @@
 import unittest
 from collections import deque
 from contextlib import _GeneratorContextManager, contextmanager, nullcontext
+from _testinternalcapi import SelfInterruptingContextManager
 
 
 def do_with(obj):
@@ -850,5 +851,21 @@ def exit_raises():
                                  expected)
 
 
+class InterruptDuringEnter(unittest.TestCase):
+
+    def test_exit_called_after_interrupt(self):
+        cm = SelfInterruptingContextManager()
+        self.assertFalse(cm.within())
+        try:
+            with cm:
+                self.assertTrue(cm.within())
+        except KeyboardInterrupt:
+            self.assertFalse(cm.within())
+            return
+        except:
+            self.fail("Wrong exception raised")
+        self.fail("No exception raised")
+
+
 if __name__ == '__main__':
     unittest.main()

Lib/test/test_with.py

@@ -284,6 +284,7 @@ DIST=		$(DISTFILES) $(DISTDIRS)
 LIBRARY=	@LIBRARY@
 LDLIBRARY=      @LDLIBRARY@
 BLDLIBRARY=     @BLDLIBRARY@
+MODULE_LDFLAGS_SHARED=$(if $(LIBPYTHON),$(BLDLIBRARY))
 PY3LIBRARY=     @PY3LIBRARY@
 DLLLIBRARY=	@DLLLIBRARY@
 LDLIBRARYDIR=   @LDLIBRARYDIR@