Enable RuntimeProvider in DoT implementations

[chengyuhui]

I was surprised that RuntimeProvider does not work in DoT, which is not ideal for my application.

Most changes of this PR are just playing around with types, almost no changes have been made to the logic.

[codecov]

Codecov Report

Merging #1373 (1fedd27) into main (6b2ed70) will increase coverage by 0.00%.
The diff coverage is 83.33%.

@@           Coverage Diff           @@
##             main    #1373   +/-   ##
=======================================
  Coverage   86.35%   86.35%           
=======================================
  Files         134      134           
  Lines       13870    13871    +1     
=======================================
+ Hits        11977    11978    +1     
  Misses       1893     1893           

[chengyuhui]

I keep getting this when running cargo make clippy locally (totally unrelated code halts the process), which is annoying.

[djc]

Are you using nightly Rust?

[chengyuhui]

Oops, my bad. Now that's sorted out.

Having to convert between std and tokio does not look great, but we can't really do something if we want to maintain async-std compatibility?

[djc]

This changes looks good in principe, though some of the type aliases look very complex. Does it really make sense to declare a TcpClientStream<AsyncIoTokioAsStd<TokioTlsStream<AsyncIoStdAsTokio<S>>>>?

In your changes, you changed the import style. I think these changes should be reverted/made to comply with the prevalent style. In the prevalent style, there is no nested grouping of imported items. I can make another pass once this is fixed.

[djc]

Let's put this on a separate line, in keeping with the surrounding style.

[chengyuhui]

Well, I don't think anyone would like such a strange type. However, I currently have no idea how to simplify that.

TcpClientStream<AsyncIoTokioAsStd<TokioTlsStream<AsyncIoStdAsTokio<S>>>> where S: Connect
^                                 ^                                ^
std                               tokio                            std                                       

The issue is that each part of this type accepts different AsyncRead + AsyncWrite coming from the standard library (or futures) and tokio, which are incompatible.

[djc]

Ugh, yeah, I guess that's actually necessary.

[chengyuhui]

I have updated all the styles.

[djc]

This looks fine to me. Thanks!

[chengyuhui]

Please note that this would be a breaking change for every TLS implementation crate, as there is an extra type parameter.