Fix crash blocking inside tokio. Fixes #1017

[ramon-garcia]

This pull request makes it possible to run blocking requests inside Tokio runtime without crashing.

This should fix issue #1017

[ramon-garcia]

Hello, please, could you allow the CI to run?

[ramon-garcia]

Hello. Any comments?

[ramon-garcia]

Ping

[seanmonstar]

Thanks for the PR!

However, this is currently by design. Blocking inside a task, even with block_in_place, can have unexpected consequences. This currently tries to warn you in development that this is a problem.

[ramon-garcia]

I followed the advice from tokio-rs/tokio#5843 . It looks like it works correctly provided that one is using the multithreaded runtime.

There are a number of cases where this is useful. For instance, one has a codebase of synchronous code, and wants to call it from an synchronous context, and converting the code does not pay off (because the web requests are really the only activity at this time).

Notice the confusing behavior, when one calls a library function from an asynchronous context, and gets a crash because some part of that library makes a web request.

What unexpected consequences could have blocking code inside block_in_place and block_on?