desktop/src-tauri/crates/core/tests/kimi_code_gaps_synthetic.rs

Comment 1, line 1

//! Kimi Code (A12, M4) boundary tests use directories and headers marked synthetic.

//! Cover error steps without usage, unregistered protocol fallback, independently verified

//! 1.4/1.5 selection, echo/completed duplicate prevention,

//! invalid negative values/scopes/second timestamps without guessed conversion, and direct detection.

//! Expected values are calculated manually in each dataset's _expectations.md.

/// Error/retry wire data without usage.record is a valid shape:

/// complete, zero events and zero diagnostics, without invented zero usage.

// No numeric source row is fabricated from absent usage.

/// Unregistered protocol "9.9" tries latest_fallback compatibility;

/// valid usage is retained as unverified, without rejecting the version number alone (V30).

/// step.end echoes equal usage.record fields; subagent.completed.usage snapshots

/// child wire sums. Neither creates events or adds duplicate usage.

// Adding echo/completed usage would inflate these four totals.

// No event may contain completed snapshot values {123,45,678,0}.

// One turn and its echo have equal sums, producing matched reconciliation.

/// Diagnose and skip negative values, unknown usageScope and second timestamps.

/// Valid records remain; do not multiply seconds by 1000. All 82 inspected native files used milliseconds.

// The only valid synthetic record {200,60,800,10} reports cache write 10;

// inspected native cache writes were zero, so this test separately checks the positive mapping.

// Second-scale time 1_767_225_600 is not converted into an event.

/// Same-millisecond usage.record entries use ordinal identity suffixes; this synthetic case was not observed locally.

// Rescanning ordinal identities adds no events.

/// Empty detection is Pending; a non-JSON or nonmetadata first line is UnknownFormat.

/// Independently verified Kimi Code protocol versions 1.5 and 1.4 both select KnownVersion.