RuntimeError: context has already been set. When I use both Sanic and loguru

[ChrisYP]

from sanic import Sanic, text
from loguru import logger

app = Sanic(__name__)


@app.get('/')
async def run_shell(request):
    return text('Hello world!')


if __name__ == '__main__':
    app.run(host='0.0.0.0', port=6666, debug=True, access_log=True, workers=2)

python version:3.8
Name: sanic
Version: 23.6.0
Name: loguru
Version: 0.7.1

The above code thorws an error:
raise RuntimeError('context has already been set')
RuntimeError: context has already been set

[Delgan]

Thanks for the report and sorry for the inconvenience.

For multiprocessing purposes, Loguru v0.7.1 added a new context argument and uses multiprocessing.get_context() by default. This is supposed to have the same effect as using multiprocessing directly, but makes it easier to switch from one context to another (eg. from "fork" to "spawn" on Linux).
Unfortunately, I didn't notice that calling multiprocessing.get_context() had the side effect of also setting the default context globally... And once a context is set, it can't be changed, hence the error you're getting.

I'll implement a fix as soon as possible. In the meantime, the only workaround I can think of is to disable the default handler of Loguru through os.environ["LOGURU_AUTOINI"] and re-add it with a context explicitly:

import os
import sys

os.environ["LOGURU_AUTOINIT"] = "0"

from sanic import Sanic, text
from loguru import logger

logger.add(sys.stderr, context="spawn")  # Assuming Sanic tries to set the start method to "spawn".

app = Sanic(__name__)


@app.get("/")
async def run_shell(request):
    return text("Hello world!")


if __name__ == "__main__":
    app.run(host="0.0.0.0", port=6666, debug=True, access_log=True, workers=2)