Backtestspydanticrun-20260923-170312#13780
Concurrent TypeAdapter construction raises RuntimeError on free-threaded Python
The merged fix was judged better than our patch.
In this judgment, Patch A = our agent, Patch B = the merged fix.
Both patches fix the exact site named in the issue (`_decorator_infos_for_class` iterating `vars(typ).items()`), and both do so with a snapshot. On correctness B's `vars(typ).copy()` is the stronger primitive: `dict.copy()` runs as a single locked C operation that is atomic against concurrent mutation, whereas Patch A's `list(vars(typ).items())` iterates the items view through the generic iterator protocol, which still carries the `ma_used` size-change check and is not clearly atomic under free-threading — so A's fix may not fully close the very window the issue describes. B also fixes the same class of race at the other iteration sites it can occur (`LazyLocalNamespace.data`), which is the same root cause (reading annotations lazily materializing `__annotations_cache__`/`__annotate_func__`). On tests, B adds a dedicated `test_freethreaded.py` that reproduces the issue's exact scenario (concurrent `TypeAdapter(TypedDict)` under a barrier) plus targeted concurrent tests for each site it touches, properly skip-guarded via a `is_free_threaded` conftest helper — these genuinely fail-before/pass-after on a free-threaded build. Patch A's regression test is single-threaded and simulates mutation via a MutatingDescriptor's `__get__`; it is doubtful the collect loop actually invokes that `__get__` during the `vars` iteration (the loop reads `.decorator_info`/`.wrapped` attributes on the proxy, not the wrapped descriptor), so the test may pass regardless of the fix — a weaker, possibly tautological test. Patch A wins on scope: it is minimal and surgical, while B's `main.py` change (replacing `delattr` with `set_model_mocks` in `model_rebuild`) is a behavioral change beyond the reported reproducer, adding risk. But that change is reasoned and covered by its own test, and scope is the third priority. Since B is stronger on the two higher-priority criteria (correctness completeness and effective, on-point regression tests), it is the better fix despite the broader footprint.
The judge saw only the two patches, labelled A and B in random order, with issue links and version numbers removed. The reasoning above is shown verbatim.
Solve
34m 13s16 turns
Wall-clock time in the container
Full test suite
No new failures
6,781 passed · 373 skipped · 7,154 total
Contamination audit
Unknown
See the audit notes for this run
Patch size
48vs 198 merged
Changed lines, ours vs merged
The issue
View on GitHubInitial Checks
- I confirm that I'm using Pydantic V2
Description
Constructing separate TypeAdapter instances concurrently for the same new TypedDict can raise RuntimeError: dictionary changed size during iteration on free-threaded Python.
A full run produced 91 failures in 800 attempts with the GIL disabled and none with PYTHON_GIL=1. It also reproduces on Pydantic main at 1c8a330acc3077e77da64e3672c7c875fb4c6eaa.
Lazy annotation evaluation adds __annotations_cache__ while another thread iterates vars(typ).items() in _decorator_infos_for_class.
Example Code
from __future__ import annotations
import threading
from concurrent.futures import ThreadPoolExecutor
from pydantic import TypeAdapter
from typing_extensions import TypedDict
def run_round() -> None:
class Record(TypedDict):
value: int
for index in range(1000):
setattr(Record, f'metadata_{index}', index)
barrier = threading.Barrier(8)
def build(_: int) -> None:
barrier.wait()
TypeAdapter(Record).validate_python({'value': '42'})
with ThreadPoolExecutor(max_workers=8) as executor:
list(executor.map(build, range(8)))
for _ in range(100):
run_round()
The synthetic attributes only widen the iteration window. They are not modified after the threads start.
Python, Pydantic & OS Version
pydantic version: 2.13.5
pydantic-core version: 2.46.5
python version: 3.14.3 free-threading build
platform: macOS-26.6.2-arm64
related packages: typing_extensions-4.15.0The two patches
pydantic/_internal/_decorators.pyindex 160a18c0..d54ee92b 100644@@ -535,7 +535,10 @@ def _decorator_infos_for_class( res = DecoratorInfos() to_replace: list[tuple[str, Any]] = [] - for var_name, var_value in vars(typ).items():+ # Take a snapshot of the class namespace: on the free-threaded build another thread may+ # mutate `typ.__dict__` (e.g. adding `__annotations_cache__` via lazy annotation evaluation)+ # while we iterate, which would raise "dictionary changed size during iteration" (see #13780).+ for var_name, var_value in list(vars(typ).items()): if isinstance(var_value, PydanticDescriptorProxy): info = var_value.decorator_info if isinstance(info, ValidatorDecoratorInfo):tests/test_decorators.pyindex 667ff813..f822cc64 100644@@ -138,3 +138,46 @@ def test_decorator_info_not_mutated() -> None: assert A.__pydantic_decorators__.computed_fields['func'].info.title is None assert B.__pydantic_decorators__.computed_fields['func'].info.title == 'test'+++def test_decorator_infos_for_class_dict_not_mutated_during_iteration():+ """Regression test for #13780.++ On the free-threaded build, one thread adding an attribute to the class+ (e.g. ``__annotations_cache__`` from lazy annotation evaluation) while another+ iterates ``vars(typ).items()`` in ``_decorator_infos_for_class`` raised+ ``RuntimeError: dictionary changed size during iteration``. Reproduce that+ deterministically by widening the class ``__dict__`` from inside the loop body.+ """+ from pydantic._internal._decorators import (+ DecoratorInfos,+ FieldValidatorDecoratorInfo,+ PydanticDescriptorProxy,+ )++ class MutatingDescriptor:+ """A descriptor whose ``__get__`` grows the owning class' ``__dict__``,+ simulating a concurrent thread adding attributes mid-iteration."""++ def __get__(self, obj, objtype=None):+ for index in range(50):+ setattr(objtype, f'_widen_{index}', index)++ def func(cls, value):+ return value++ return func++ class Model:+ pass++ Model.my_validator = PydanticDescriptorProxy(+ wrapped=MutatingDescriptor(),+ decorator_info=FieldValidatorDecoratorInfo(+ fields=('x',), mode='after', check_fields=None, json_schema_input_type=None+ ),+ )++ decorators = DecoratorInfos.build(Model)++ assert 'my_validator' in decorators.field_validators
pydantic/_internal/_decorators.pyindex 160a18c00..72b668c53 100644@@ -535,7 +535,12 @@ def _decorator_infos_for_class( res = DecoratorInfos() to_replace: list[tuple[str, Any]] = [] - for var_name, var_value in vars(typ).items():+ # Iterate on the vars with a copy to avoid changes in dictionary size during iteration.+ # This can happen if a concurrent thread has Pydantic processing the same `typ` (e.g. two+ # concurrent threads doing `TypeAdapter(SomeClassType)`), and such thread reads annotations+ # for the first time during the iteration here in the other thread (since 3.14, reading annotations+ # can set attributes such as `__annotations_cache__` or `__annotate_func__` to the type).+ for var_name, var_value in vars(typ).copy().items(): if isinstance(var_value, PydanticDescriptorProxy): info = var_value.decorator_info if isinstance(info, ValidatorDecoratorInfo):pydantic/_internal/_namespace_utils.pyindex d308f7e4c..495f17da6 100644@@ -4,6 +4,7 @@ import sys from collections.abc import Callable, Generator, Iterator, Mapping from contextlib import contextmanager from functools import cached_property+from types import MappingProxyType from typing import Any, NamedTuple, TypeAlias, TypeVar from typing_extensions import ParamSpec, TypeAliasType, TypeVarTuple@@ -88,7 +89,12 @@ class LazyLocalNamespace(Mapping[str, Any]): @cached_property def data(self) -> dict[str, Any]:- return {k: v for ns in self._namespaces for k, v in ns.items()}+ # In rare scenarios, a race condition can happen when iterating over class namespaces (i.e. created with `var(SomeClass)`),+ # if another thread accesses the type annotations for the first time (since 3.14, reading annotations can set attributes+ # such as `__annotations_cache__` or `__annotate_func__` to the type).+ return {+ k: v for ns in self._namespaces for k, v in (ns.copy() if isinstance(ns, MappingProxyType) else ns).items()+ } def __len__(self) -> int: return len(self.data)pydantic/main.pyindex 1c0258c69..395c3c85f 100644@@ -711,16 +711,12 @@ class BaseModel(metaclass=_model_construction.ModelMetaclass): cls.__pydantic_complete__ = False - for attr in ('__pydantic_core_schema__', '__pydantic_validator__', '__pydantic_serializer__'):- if attr in cls.__dict__ and not isinstance(- getattr(cls, attr), (_mock_val_ser.MockCoreSchema, _mock_val_ser.MockValSer)- ):- # Deleting the validator/serializer is necessary as otherwise they can get reused in- # pydantic-core. Same applies for the core schema that can be reused in schema generation.- # We do so only if they aren't mock instances, otherwise concurrent reads of these attributes- # — performed without holding the rebuild lock (e.g. when instantiating the model) — can- # resolve them from the parent class.- delattr(cls, attr)+ if already_complete:+ # The existing validator/serializer must not be reused in pydantic-core, and the core schema+ # must not be reused in schema generation. If the model is already complete (i.e. `force=True`),+ # we set back mocks for the model (we could also delete the complete core schema/validator/serializer,+ # but this isn't thread-safe).+ _mock_val_ser.set_model_mocks(cls) if _types_namespace is not None: rebuild_ns = _types_namespacetests/conftest.pyindex b066a013a..694c5855b 100644@@ -23,6 +23,11 @@ from pydantic import BaseModel from pydantic._internal._generate_schema import GenerateSchema from pydantic.json_schema import GenerateJsonSchema +try:+ is_free_threaded = not sys._is_gil_enabled()+except AttributeError:+ is_free_threaded = False+ def pytest_addoption(parser: pytest.Parser): parser.addoption('--test-mypy', action='store_true', help='run mypy tests')tests/test_freethreaded.pynew file mode 100644index 000000000..af4ef4929@@ -0,0 +1,162 @@+import sys+import threading+from collections.abc import Callable+from concurrent.futures import ThreadPoolExecutor+from typing import Any++import pytest+from typing_extensions import TypedDict++import pydantic.dataclasses+from pydantic import BaseModel, TypeAdapter, field_validator++from .conftest import is_free_threaded++pytestmark = [+ pytest.mark.skipif(sys.platform == 'emscripten', reason='no threading on emscripten'),+ pytest.mark.skipif(not is_free_threaded, reason='only relevant on free-threaded builds'),+]+++def run_concurrently(+ setup: Callable[[], Any], target: Callable[[Any, int], Any], *, rounds: int = 100, threads: int = 8+):+ """Run `target` in `threads` threads at once, `rounds` times, with a fresh `setup()` result each round."""+ for _ in range(rounds):+ ctx = setup()+ barrier = threading.Barrier(threads)++ def work(index: int) -> None:+ barrier.wait()+ target(ctx, index)++ with ThreadPoolExecutor(max_workers=threads) as executor:+ for future in [executor.submit(work, i) for i in range(threads)]:+ future.result()+++def test_typed_dict_concurrent_type_adapter() -> None:+ def setup():+ class Record(TypedDict):+ value: int++ return Record++ def target(Record, _):+ assert TypeAdapter(Record).validate_python({'value': '42'}) == {'value': 42}++ run_concurrently(setup, target)+++def test_base_model_concurrent_creation_with_shared_base() -> None:+ def setup():+ class Mixin:+ a: int = 1++ @field_validator('a')+ @classmethod+ def val_a_1(cls, v: int) -> int:+ return v++ @field_validator('a')+ @classmethod+ def val_a_2(cls, v: int) -> int:+ return v++ @field_validator('a')+ @classmethod+ def val_a_3(cls, v: int) -> int:+ return v++ return Mixin++ def target(Mixin, _):+ class Model(BaseModel, Mixin):+ b: int++ assert Model(b='1').b == 1++ run_concurrently(setup, target)+++def test_dataclass_concurrent_creation_with_shared_base() -> None:+ def setup():+ class Base:+ a: int = 1++ @field_validator('a')+ @classmethod+ def val_a_1(cls, v: int) -> int:+ return v++ @field_validator('a')+ @classmethod+ def val_a_2(cls, v: int) -> int:+ return v++ @field_validator('a')+ @classmethod+ def val_a_3(cls, v: int) -> int:+ return v++ return Base++ def target(Base, _):+ @pydantic.dataclasses.dataclass+ class Sub(Base):+ b: int++ assert Sub(b='1').b == 1++ run_concurrently(setup, target)+++def test_model_rebuild_concurrent_subclass_creation(monkeypatch: pytest.MonkeyPatch) -> None:+ """Test targeted at our `LazyLocalNamespace.data` property implementation."""++ def setup():+ monkeypatch.delitem(globals(), 'Later', raising=False)++ class Model(BaseModel):+ # `Later` isn't defined yet, so the field is deferred and no annotations cache is set on the class:+ x: Later # noqa: F821 # pyright: ignore[reportUndefinedVariable]++ # Widen the class dict so that iterating over it takes long enough for the race to happen:+ for i in range(3000):+ setattr(Model, f'attr_{i}', i)++ # Dynamically set the item in globals instead of a simple assignment to force `ForwardRef.evaluate()` into looking at the+ # provided namespace during evaluation (otherwise it doesn't look at globals/locals and uses `ForwardRef.__cell__` directly)+ monkeypatch.setitem(globals(), 'Later', int)+ return Model++ def target(Model, index: int):+ if index == 0:+ # Collecting the fields of the subclass reads the annotations of `Model` for the first time:+ class Sub(Model):+ pass++ assert Sub(x='1').x == 1+ else:+ Model.model_rebuild(force=True)++ run_concurrently(setup, target)+++def test_model_force_rebuild_concurrent_instantiation() -> None:+ """Test targeted at our logic in `model_rebuild()`, when we set back mocks on already complete models."""++ def setup():+ class Model(BaseModel):+ x: int++ assert Model(x=1).x == 1 # `Model` is complete, with a real validator set on the class+ return Model++ def target(Model, index):+ if index == 0:+ Model.model_rebuild(force=True)+ else:+ assert Model(x='1').x == 1++ run_concurrently(setup, target)
What the agent reported
The review confirms the fix is correct, minimal, and complete. Here's my final report.
Summary
What was wrong: In pydantic/_internal/_decorators.py, _decorator_infos_for_class iterated the class namespace directly with for var_name, var_value in vars(typ).items():. On free-threaded Python, when separate TypeAdapter instances are built concurrently for the same new TypedDict, one thread's lazy annotation evaluation adds __annotations_cache__ to the class dict while another thread is mid-iteration over it, raising RuntimeError: dictionary changed size during iteration (GitHub #13780).
What I changed:
pydantic/_internal/_decorators.py(line 538): snapshot the namespace before iterating —for var_name, var_value in list(vars(typ).items()):— with an explanatory comment referencing #13780. This is the minimal fix; the loop body is untouched. The function is the single collection path for bothtypand its bases, so both are covered.tests/test_decorators.py: addedtest_decorator_infos_for_class_dict_not_mutated_during_iteration, a deterministic regression test. It places aPydanticDescriptorProxywrapping a descriptor whose__get__widens the class__dict__;DecoratorInfos.buildreaches it mid-loop, mutating the dict exactly as the free-threaded race does. This reproduces the identicalRuntimeErroron a normal GIL build (the reporter's threaded snippet only fails on free-threaded builds, so it could not serve as a gate).
Test evidence:
- Before the fix: the new test failed with
RuntimeError: dictionary changed size during iterationat_decorators.py:538— the exact reported error. - After the fix:
./bt-test tests/test_decorators.py→ 32 passed (including the new test);./bt-test tests/test_validators.py(heavy consumer of this path) → 179 passed, 2 pre-existing xfailed. No regressions. - Reviewer confirmed the diff is correct, minimal, touches only the two files, and adds no scratch files or weakened tests.
Requirements list:
- Fix concurrent
TypeAdapterconstruction raisingRuntimeErroron free-threaded Python (the sole ask / "Initial Checks") — implemented (snapshot fix) with test evidence (fail-before/pass-after above). - Example reproduction scenario (concurrent
TypeAdapter(Record)over a mutatedTypedDict) — covered by a deterministic regression test exercising the same_decorator_infos_for_classiteration and asserting the build succeeds; verified failing before and passing after.
All changes are left uncommitted in the working tree.