Observability

src/logging.rs

EchoWarrior uses tracing for structured logging.

logging::init() installs:

• a stderr formatting layer controlled by RUST_LOG
• a custom ring-buffer layer that stores the last 50 formatted events

The returned LogBuffer is cheap to clone and is passed into the runtime so the F1 debug overlay can show recent messages.

Log Buffer Behavior

LogBuffer stores lines in a VecDeque behind Arc<Mutex<_>>.

Important details:

• capacity is LOG_CAPACITY = 50
• snapshot() returns oldest-to-newest lines
• last() returns the newest line
• poisoned locks return empty/no value instead of panicking
• initializing logging twice prints a stderr notice but does not crash

This follows the project-wide graceful degradation rule: debug tooling must not take down the game.

src/perf.rs

src/perf.rs provides debug-only timing scopes:

perf_scope!("runtime.update");
perf_scope!("asset_pack.discovery", 12.0);

In debug builds, a PerfScope logs a throttled warning through tracing when elapsed time exceeds the budget. In release builds, the same API compiles to no work.

The default budget is 4 ms and can be changed with:

$env:ECHO_WARRIOR_PERF_WARN_MS = "8"

Warnings use the echo_warrior::perf target, so they appear in stderr and the F1 debug overlay.

When To Add A Scope

Add perf_scope! around expensive debug paths such as:

• asset loading
• pack discovery or verification
• save load/write
• Lua dispatch
• weather/render preparation
• pathfinding or broadphase code

Keep scopes coarse enough to explain a slowdown without flooding logs.