fix(multipath): fix remote state actor termination #3676
Conversation
|
Documentation for this PR has been generated and is available at: https://n0-computer.github.io/iroh/pr/3676/docs/iroh/ Last updated: 2025-11-20T09:37:05Z |
Frando
force-pushed
the
Frando/mp-fix-remote-state-actor-loop
branch
from
6ffcbb0 to
7c11665
Compare
| let task = task::spawn( | ||
| async move { | ||
| if let Err(err) = self.run(rx).await { | ||
| error!("actor failed: {err:#}"); |
There was a problem hiding this comment.
So this was a poor-man's attempt at being able to have "fatal" errors. The right thing of course is to panic and spawn the tasks in a JoinSet and making sure the panic bubbles up to the endpoint. But that's for another day I guess. For now making this infallible is the safer approach I think. While if forces you to handle the fatal error everywhere it will probably result in better code.
iroh/src/magicsock/transports.rs
Outdated
| if any_match { | ||
| Err(io::Error::other("all available transports failed")) | ||
| Err(io::Error::other(format!( | ||
| "all available transports failed for destination {dst:?}" |
There was a problem hiding this comment.
Hum, convention in the stdlib is that the caller should provide the context. Don't we have somewhere else to add context?
There was a problem hiding this comment.
sgtm. pushed a commit that does this.
…send_datagram` (#3678) Intended for merging into #3676 - Makes `idle_timeout` always be a `time::Sleep` that is `.reset` once there is any activity. I find this much easier to reason about. - Fixes a bug where `handle_msg_send_datagram` short-circuits when sending on one path fails. Instead, it should *try* all paths, even if some of them fail.
…ix-remote-state-actor-loop
Description
Breaking Changes
Notes & open questions
Change checklist
quic-rpciroh-gossipiroh-blobsdumbpipesendme