First Check

• I added a very descriptive title to this issue.
• I used the GitHub search to find a similar issue and didn't find it.
• I searched the FastAPI documentation, with the integrated search.
• I already searched in Google "How to X in FastAPI" and didn't find any information.
• I already read and followed all the tutorial in the docs and didn't find an answer.
• I already checked if it is not related to FastAPI but to Pydantic.
• I already checked if it is not related to FastAPI but to Swagger UI.
• I already checked if it is not related to FastAPI but to ReDoc.

Commit to Help

• I commit to help with one of those options 👆

Example Code

import asyncio

import uvicorn
from fastapi import FastAPI
from fastapi.responses import StreamingResponse

app = FastAPI()


def sync_streamer():
    try:
        while True:
            yield b"--boundary\r\nContent-Type: text/plain\r\nContent-Length: 1\r\n\r\n1\r\n"
    except asyncio.CancelledError:
        print("caught cancelled error")


async def async_streamer():
    try:
        while True:
            yield b"--boundary\r\nContent-Type: text/plain\r\nContent-Length: 1\r\n\r\n1\r\n"
    except asyncio.CancelledError:
        print("caught cancelled error")


@app.get("/sync")
async def sync_endpoint():
    return StreamingResponse(
        sync_streamer(),
        media_type="multipart/x-mixed-replace; boundary=boundary",
    )


@app.get("/async")
async def async_endpoint():
    return StreamingResponse(
        async_streamer(),
        media_type="multipart/x-mixed-replace; boundary=boundary",
    )

Description

When I try to send data with a StreamingResponse from an infinite generator I sometimes get an error when the generator is asynchronous. Sometimes it works fine, but sometimes I get an infinite stream of errors: socket.send() raised exception. on the server side.
If I instead use a synchronous generator this problem does not appear at all.

I've based my sample code on #4146 and the CancelledError does get caught sometimes, but sometimes it doesn't. Right now I can avoid this issue by making the generator synchronous, but it would be nice to have the option to use async generators as well.

It could be that this is a Starlette issue but I am really not sure.

Operating System

Linux

Operating System Details

No response

FastAPI Version

0.79.0

Python Version

3.8.10

Additional Context

No response