*: Remove use of development_transport in examples and tests
#3056
Conversation
These create circular dependencies to the root libp2p crate in tests.
| MyBehaviour { | ||
| rendezvous: rendezvous::client::Behaviour::new(identity.clone()), | ||
| ping: ping::Behaviour::new(ping::Config::new().with_interval(Duration::from_secs(1))), | ||
| }, | ||
| PeerId::from(identity.public()), | ||
| ); | ||
| ) | ||
| .executor(Box::new(|f| { |
There was a problem hiding this comment.
Curious Thomas, why do we need it here but not when using tcp::async_io::Tcp ?
There was a problem hiding this comment.
async-std implicitly creates a reactor on each thread that an IO component is polled on. tokio only creates a reactor on worker threads and thus requires explicit control over where tasks are spawned.
By default, a Swarm creates a futures::executor::Threadpool for these background tasks and that doesn't work for tokio. Also see #3068.
|
just noticed https://github.com/libp2p/rust-libp2p/blob/master/src/tutorials/ping.rs also needs to be addressed |
|
How should we move forward here? Should I remove the deprecation and just not use those transports for the tests and crate-specific examples? That will break the circular dependencies which is what I am after. We can then decide separately, what we should rename those transport constructors to. @mxinden Input please :) |
thomaseizinger
changed the title
development_transport in examples and tests
|
Closing in favor of #3023. |
Description
These create circular dependencies to the root libp2p crate in tests.
Links to any relevant issues
libp2p-swarm-test#2888: Once we have the shared test abstraction, more duplication in test setup will go away again. For the examples, I think we should not be using thedevelopment_transportbut be as close to a real-world application as possible.development_transportinside our workspace.Open Questions
Change checklist
I have added tests that prove my fix is effective or that my feature works