MagicStack / MagicStack/uvloop

uvloop debug mode crashes on Python 3.14 during async generator finalization (traceback.walk_stack sees non-frame object)

Abierto
#715 4 comentarios 2 reacciones 0 asignados Ver en GitHub

Nadie ha tomado este issue todavía.

Lenguaje dominante
Cython
Estrellas
11.9k
Forks
615
Métricas de merge de PR
Sin PR fusionados en 30 d

Descripción

Overview

When an async generator is finalized under uvloop debug mode, uvloop’s extract_stack() calls traceback.walk_stack() on Python 3.14. The stack walker receives a non-frame object (e.g. _asyncio.TaskStepMethWrapper) and crashes (traceback.walk_stack_generator) or raises AttributeError. It's worth mentioning this does NOT happen with asyncio debug.

Environment

  • uvloop 0.22.1
  • Python 3.14.0
  • macOS (Darwin 24.6.0, arm64)
  • Debug mode enabled (PYTHONASYNCIODEBUG=1 or uvloop.run(debug=True))

Reproduction

  1. Set PYTHONFAULTHANDLER=1 PYTHONASYNCIODEBUG=1 as environment variables.
  2. Run the following script with python3 uvloop_python314_bug_repro.py
"""uvloop_python314_bug_repro.py

Minimal repro: uvloop 0.22.1 + Python 3.14 debug async-gen finalization crash.

This creates a nested async-generator chain and breaks on a "[DONE]" sentinel
without closing the inner iterator. Under uvloop debug, async generator
finalization schedules aclose() via call_soon_threadsafe, which calls
extract_stack(). On Python 3.14, traceback.walk_stack sees a non-frame object
(for example, _asyncio.TaskStepMethWrapper) and raises AttributeError. The
unpatched build can segfault in traceback.walk_stack_generator.

Run:
  uv run --active --no-sync python uvloop_python314_bug_repro.py
"""

from __future__ import annotations

import asyncio
import gc
import os
import sys
from typing import AsyncIterator

import uvloop

os.environ["PYTHONASYNCIODEBUG"] = "1"

REPEATS = 1


class FakeResponse:
    async def aiter_raw(self) -> AsyncIterator[bytes]:
        payloads = [
            b"data: one\n\n",
            b"data: [DONE]\n\n",
            b"data: trailing\n\n",
        ]
        for payload in payloads:
            yield payload
            await asyncio.sleep(0)

    async def aclose(self) -> None:
        await asyncio.sleep(0)


class Decoder:
    async def _aiter_chunks(self, source: AsyncIterator[bytes]) -> AsyncIterator[bytes]:
        async for chunk in source:
            yield chunk

    async def aiter_bytes(self, source: AsyncIterator[bytes]) -> AsyncIterator[str]:
        async for chunk in self._aiter_chunks(source):
            for line in chunk.splitlines():
                yield line.decode("utf-8")


class AsyncStream:
    def __init__(self) -> None:
        self.response = FakeResponse()
        self.decoder = Decoder()
        self._iterator: AsyncIterator[str] = self.__stream__()

    def __aiter__(self) -> AsyncIterator[str]:
        return self._iterator

    async def __stream__(self) -> AsyncIterator[str]:
        iterator = self._iter_events()
        try:
            async for event in iterator:
                if event.endswith("[DONE]"):
                    break
                yield event
        finally:
            await self.response.aclose()

    async def _iter_events(self) -> AsyncIterator[str]:
        async for event in self.decoder.aiter_bytes(self.response.aiter_raw()):
            yield event


async def exercise_once() -> None:
    stream = AsyncStream()
    async for _ in stream:
        pass
    del stream
    gc.collect()
    await asyncio.sleep(0)


async def main() -> None:
    loop = asyncio.get_running_loop()
    print(f"Python: {sys.version}")
    print(f"Loop: {type(loop).__name__}")
    print(f"Debug: {loop.get_debug()}")
    for _ in range(REPEATS):
        await exercise_once()
    await asyncio.sleep(0.1)


if __name__ == "__main__":
    print(f"uvloop: {uvloop.__version__}")
    uvloop.run(main(), debug=True)

Expected

Clean exit (debug stack capture should never crash the process).

Actual

Fatal Python error: Segmentation fault
File ".../traceback.py", line 393 in walk_stack_generator
...

Ideas

Python 3.14 changed frame introspection behavior. During async generator finalization, uvloop's debug stack capture (cbhandles.pyx:extract_stack) calls traceback.walk_stack(). The stack can contain non-frame objects like _asyncio.TaskStepMethWrapper which lack the f_back attribute that walk_stack_generator expects.

The crash path:

  1. Async generator breaks early (e.g., on [DONE] sentinel)
  2. GC finalizes the generator
  3. Finalizer schedules aclose() via call_soon_threadsafe
  4. uvloop debug captures stack → traceback.walk_stack() → crash

This happens during async generator finalization; the finalizer schedules aclose() via call_soon_threadsafe, which triggers uvloop’s debug stack capture. A defensive guard before calling traceback.walk_stack (e.g. require a real frame, or swallow AttributeError/TypeError) avoids the crash.

Workaround

Explicitly close/break async generator chains before they go out of scope, or disable uvloop debug mode on Python 3.14.

Guía de contribución

No hay ninguna guía de contribución indexada para este repositorio

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Línea de trabajo

Comienza con cbhandles.pyx:extract_stack y el comportamiento de traceback.walk_stack en Python 3.14; después reproduce el fallo usando uvloop_python314_bug_repro.py con el modo de depuración habilitado. Sigue la finalización de los generadores asíncronos a través de call_soon_threadsafe y verifica que la reproducción termine correctamente sin un AttributeError ni un fallo de segmentación.

Escrito por el modelo de indexación a partir del texto del issue.

Evaluación

Stack tecnológico
python
Área
backend
Tipo de issue
Error
Dificultad
4/5
Tiempo estimado
3-5 días
Estado de actividad
Tranquilo
Claridad
Bastante claro
Aptitud para principiantes
48/100

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.