为何在done回调中调用future.result()会改变异常回溯信息？

问题现象 #

当同时满足以下三个条件时，CPython环境下asyncio协程抛出的异常回溯会丢失内部调用栈：

• 给协程对应的Future对象绑定了调用fut.result()的done回调
• 在其他代码位置await该Future
• 协程内部抛出异常

移除done回调里的fut.result()调用后，就能得到包含协程内部调用栈的完整回溯；而在PyPy环境中，无论是否调用fut.result()，都能正常显示完整回溯。

复现代码 #

import asyncio

async def my_traceback_gets_mangled():
    def bar():
        raise TypeError('I wish I had a traceback that let me find this bogus raise')
    await asyncio.sleep(1)
    bar()

def done_cb(fut):
    try: fut.result()
    except: pass
    
async def main():
    fut = asyncio.ensure_future(my_traceback_gets_mangled())
    fut.add_done_callback(done_cb)
    await fut

asyncio.run(main())

输出对比 #

调用fut.result()时的截断回溯 #

Traceback (most recent call last):
  File "/tmp/foo.py", line 18, in <module>
    asyncio.run(main())
  File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
    return future.result()
  File "/tmp/foo.py", line 16, in main
    await fut
TypeError: I wish I had a traceback that let me find this bogus raise

移除fut.result()后的完整回溯 #

Traceback (most recent call last):
  File "/tmp/foo.py", line 18, in <module>
    asyncio.run(main())
  File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
    return future.result()
  File "/tmp/foo.py", line 16, in main
    await fut
  File "/tmp/foo.py", line 12, in my_traceback_gets_mangled
    bar()
  File "/tmp/foo.py", line 10, in bar
    raise TypeError('I wish I had a traceback that let me find this bogus raise')
TypeError: I wish I had a traceback that let me find this bogus raise

原因 #

这是CPython中asyncio Future的实现细节：第一次调用fut.result()时，会修改异常对象的回溯信息，只保留到result()调用点的栈轨迹。后续通过await触发异常抛出时，会复用这个已经被修改的异常对象，导致协程内部的原始调用栈丢失。

PyPy的asyncio实现没有这个回溯截断逻辑，所以能保留完整调用栈。

解决办法 #

1. 取消done回调中的fut.result()调用 #

如果done回调不需要处理异常，直接删掉fut.result()相关代码即可，这样await时会抛出带完整回溯的异常。

2. 用fut.exception()替代fut.result()获取异常 #

如果需要在done回调中处理异常，使用fut.exception()获取异常对象，避免触发result()的回溯修改逻辑：

def done_cb(fut):
    exc = fut.exception()
    if exc is not None:
        # 在这里处理异常，比如打印日志
        print(f"捕获异常: {exc}")

3. 手动保留并恢复原始回溯 #

如果必须调用fut.result()，可以先保存原始异常的回溯信息，之后再恢复：

def done_cb(fut):
    exc = fut.exception()
    original_traceback = exc.__traceback__ if exc else None
    try:
        fut.result()
    except:
        pass
    # 恢复原始回溯，确保后续await能拿到完整栈
    if exc:
        exc.__traceback__ = original_traceback

内容的提问来源于stack exchange，提问作者Sam Hartman