eventlet.websocket is not always used from eventlet.wsgi, so do not assume eventlet.set_idle exists

[dhdaines]

In the case where Eventlet's WebSocket implementation is used, for instance, in a Gunicorn worker, it ends up being the case that WebSocketWSGI is not running under an Eventlet server, and thus shouldn't assume that the environment contains eventlet.set_idle (which is an implementation detail of the Eventlet server and thus quite meaniningless).

This is notably the case when using Flask-SocketIO with Gunicorn as suggested in the documentation: https://flask-socketio.readthedocs.io/en/latest/deployment.html#gunicorn-web-server

It seems quite safe to just avoid trying to call it in this case. If, for instance, the engineio wrapper (https://github.com/miguelgrinberg/python-engineio/blob/main/src/engineio/async_drivers/eventlet.py#L39) wishes to provide eventlet.set_idle for some reason, then it could do so.

Fixes #946

[dhdaines]

Note that I don't know of any good way to test this inside Eventlet... it seems fiendishly complicated to mock a Gunicorn worker, so that's left as an exercise to the maintainer :)

[4383]

Concerning myself, this patch LGTM.

I'd appreciate to have a secondary opinion from Tim.

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 56%. Comparing base (80f3936) to head (2082573).

Files	Patch %	Lines
eventlet/websocket.py	50%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@          Coverage Diff          @@
##           master   #949   +/-   ##
=====================================
- Coverage      56%    56%   -1%     
=====================================
  Files          89     89           
  Lines        9766   9767    +1     
  Branches     1818   1819    +1     
=====================================
  Hits         5475   5475           
- Misses       3920   3921    +1     
  Partials      371    371           

Flag	Coverage Δ
ipv6	23% <0%> (-1%)	⬇️
py310asyncio	52% <50%> (-1%)	⬇️
py310epolls	53% <50%> (-1%)	⬇️
py310poll	53% <50%> (-1%)	⬇️
py310selects	53% <50%> (-1%)	⬇️
py311asyncio	52% <50%> (-1%)	⬇️
py311epolls	53% <50%> (-1%)	⬇️
py312asyncio	50% <50%> (-1%)	⬇️
py312epolls	51% <50%> (-1%)	⬇️
py37asyncio	50% <50%> (-1%)	⬇️
py37epolls	51% <50%> (-1%)	⬇️
py38asyncio	51% <50%> (-1%)	⬇️
py38epolls	53% <50%> (-1%)	⬇️
py38openssl	51% <50%> (-1%)	⬇️
py38poll	53% <50%> (-1%)	⬇️
py38selects	53% <50%> (+<1%)	⬆️
py39asyncio	51% <50%> (-1%)	⬇️
py39dnspython1	51% <50%> (-1%)	⬇️
py39epolls	53% <50%> (+<1%)	⬆️
py39poll	53% <50%> (-1%)	⬇️
py39selects	52% <50%> (+<1%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[4383]

Well, I think these changes are well isolated and should not trigger side effects. Let's go ahead to fix the latest release.
I'm going to prepare a new release.

Thanks @dhdaines for your patch,for your time and for the interesting discussion. Much appreciated.