Sergey Shepelev

@AGSPhoenix currently, eventlet backdoor basically patches stdin/stdout to socket and runs stdlib `code.InteractiveConsole`. Personally, I approve implementing REPLoop in backdoor - it's mostly copy code from stdlib and removing sys.stdin/out...

> Maybe the lost lines are somehow being eaten when eventlet switches stdin around? My guess this is possible. > Maybe this is some weird quirk of how eventlet wraps...

@AGSPhoenix what questions? - it changes `sys.stdout` - you can remove `switch()` - working with file objects fixes need for `SocketStdout` wrapper, access to sturdy `readline` implementation, and just a...

Thank you, I'll fix it.

@rawrrawrfox thank you. Configurable location of resolv.conf is a great feature. I think we should try to support one of: - `EVENTLET_RESOLV_CONF=/path/to/resolv.conf` or just `RESOLV_CONF` at first, becase second option...

@m0ppers thanks, I'll try to fix it sooner. `socket=True` selects to patch only that module, so `ssl` is still blocking. That would heavily restrict possible improvements from eventlet anyway.

@shettyritesh if that's a possible solution for you, patch multiprocessing stdlib or copy it into your virtualenv with something along these lines: **WARNING: untested dirty hack ahead** ``` --- a/connection.py...

Patch above is not an "official recommended solution", but rather a quick dirty workaround in case you really need it fast.

Yes, it was supposed to cure that exact error. It seemed like a viable idea, though I didn't test it. I'm sorry, can't help more today. Please, wait for working...

@tuxmaster5000 Sorry, our `setup.py` is designed for build/install only. Should be documented. I understand `setup.py test` is a decent API that is nice to have working, but it doesn't here....