Eric

> Maybe I wasn't clear, but the idea would be to keep the callback, just call it once at the end. Pseudocode: > > ```nim > proc start*(s: Switch) =...

This PR now has: 1. support for `switch.listenError` callback 2. support for `transport.listenError` callback, unique *per transport* 3. tests that have fully and partially optimistic/pessimistic cases 4. tests for changes...

`switch.listenError` has been removed, along with filtering of addresses in transport's base `handles`.

> @emizzle we don't need to mock connections, instead we use the BufferStream abstraction for that. Please take a look at `test/helpers.nim` where a wrapper `TestBufferStream` is defined which allows...

> I'm still a bit conflicted about how we're approaching this from an API standpoint. > > In our last call @Menduist brought up a good point about passing the...

> The second point that @Menduist brought up was in regards to [#668 (comment)](https://github.com/status-im/nim-libp2p/pull/668#discussion_r765708049), which allows passing the callback to `start` instead the constructor. I don't have a strong opinion...

[Latest commit](https://github.com/status-im/nim-libp2p/pull/668/commits/d4a65503ac1dbd89e208730261e29671248367e1) adds two things: 1. During switch instantiation, raises defect if no transports or addresses supplied 2. During transport.start, raises defect if no addresses supplied. I implemented the points...

Anyone know why CI is failing for `testnative` on Linux? I've run this in a Ubuntu 20.04.3 VM using nim `1.4.8` and `1.2.16` and had no failures...

All tests passing.

> You can also add a link to your tutorial at the end of the second one, and add it to the CI in libp2p.nimble Fixed up. Also changed the...