Ash Berlin-Taylor

> Thanks for your contribution! > > Monkeypatching `sys.stdout` feels quite risky though. A few questions: > - What will happen on Windows terminals that use an encoding incompatible with...

So adding a `int(platform..release()) > 10` sounds like a safer way of doing this then? ```python console >>> platform.uname() uname_result(system='Windows', node='sinope', release='10', version='10.0.22621', machine='AMD64') >>> platform.release() '10' ```

Ah I've done a bit of digging and technically it's not even all versions of windows 10, but anyone since the Creators update, build 1809, so I'll make the check...

Any other comments or changes requested @ambv @JelleZijlstra?

If we don't accept this (and I entirely understand why) is there perhaps a place in the docs that I could document the work around to users (setting `PYTHONUTF8=1`)?

One comment on the "We're monkeypatching the standard library, always a risky thing to do" -- if that is the only concern then I could replace the `sys.stdout = ...`...

This may not be a problem, and all I want is an option to _not_ warn on imports that can't be resolved from outside "airflow."

@ruarfff Can you let me know your Astro support ticket number and we'll dig in to this a bit deeper. Edit: found the ticket now. Okay, saying "They couldn't figure...

Yeah I get that! S'alright. We might be in touch if we need help reproducing this.

I'd really rather we didn't run R in the airflow unit test suite - that isn't our job.