Struct CallEvents 

pub struct CallEvents { /* private fields */ }

A call’s event stream: one receiver, bounded, owned by whoever calls Call::events first.

There is exactly one consumer by construction (vision principle 3 — own it, don’t share it): Call::events hands this out once and returns None on every call after, rather than a value anyone could clone a second reader from.

Implementations

impl CallEvents

pub fn dropped(&self) -> u64

How many of this call’s events were dropped because this consumer was behind (§12.1).

Per call, and reported to the consumer that lost them, rather than folded into an endpoint-wide total. The overflow policy above is deliberate — a slow consumer loses history, not correctness — but §12.1’s rule is that a discard is counted, and until X-54 this one was reported with a tracing::debug! and nothing else, which is the exact failure that section names.

It is not in SignallingCounts because a crate-wide total would say that some call somewhere lost some events, which is not something anyone can act on. The party who can act is the one holding this receiver: it fell behind, and resynchronising from the next event’s snapshot is the documented recovery (docs/specs/app-contract.md §5.1). Monotonic, and never reset by reading it.

impl CallEvents

pub async fn recv(&mut self) -> Option<CallEvent>

The next event, or None once the call has dropped its sender.

A well-behaved consumer only ever sees None after it has already seen CallEvent::Ended — the sender is not dropped until the call’s own destructor runs, which is after the last event has been queued.

pub fn try_recv(&mut self) -> Option<CallEvent>

The next event if one is already queued, without waiting.

Trait Implementations

impl Debug for CallEvents

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.