nat-goodspeed

I wonder if the author encountered #483? For a nontrivial child-process run, I *want* the pipes passed to that child to be nonblocking.

The first step is to verify that we already have at least one test that demonstrates this issue with Python 3.7.3. If not, such a test must be added. That's...

Well. Travis's python37 is 3.7.1, which doesn't drive the problem.

Okay, so [I specified Python 3.7.3](https://github.com/eventlet/eventlet/pull/565/commits/e9cc890429fc4b96b7d913134280a08d3259cfa5) for Travis testing -- and [all existing tests passed](https://travis-ci.org/eventlet/eventlet/builds/528173659?utm_source=github_status&utm_medium=notification)! Yet I can reproduce #526 merely by attempting a simple `requests` HTTPS connection from my...

Odd. The new test script is currently failing to [import requests](https://travis-ci.org/eventlet/eventlet/jobs/587562973#L1446) -- despite the successful [pip install requests](https://travis-ci.org/eventlet/eventlet/jobs/587562973#L323) earlier in the same run. I assume that failure is typical of...

Yes: I just clicked through all the reported failures. Same thing: missing `requests` module.

Good news: we finally have a test that drives the problem! Bad news: I inadvertently broke another test. Working on that.

Okay! On this branch we have a single new test that drives the reported problem.

That's true; but for nonblocking I/O to a Windows child process, use of a named pipe with a generated name is more typical.

I encountered a different odd interaction of `monkey_patch()` options on macOS 10.12.6: if you pass `builtins=True` and `subprocess=True`, you *must* also pass `os=True`. With only `builtins=True` and `subprocess=True`: $ python...