Enum Coverage 

pub enum Coverage {
    Normalized,
    Retained,
    Dropped,
    Unmodeled,
}

How a given discriminant is handled by the loader.

Variants

Normalized

Parsed and normalized into the canonical conversation.

Retained

PARITY-12/PARITY-13 (P012/P013): parsed and its content/provenance IS captured by the loader — into Session.meta (Codex session_meta’s id/cwd/model/base_instructions, turn_context’s model), replay semantics (thread_rolled_back actually removes the rolled-back turns, exited_review_mode’s review_output.overall_explanation becomes a message with review_output.findings AND overall_correctness/overall_confidence_score (N4) captured onto that message’s metadata (D4/N4), thread_goal_updated‘s goal.objective becomes a message with goal.status/ goal.tokenBudget captured onto that message’s metadata too (D4), agent_message can become a message when its text has no response_item twin) — just not 1:1 into a ChatMessage the way Normalized records are. This is the “retained, not dropped” bucket the two audits were missing: before this variant existed, every one of these landed in Dropped indistinguishably from truly-inert UI noise (token_count, task_started, …), which is exactly the false “silently dropped” signal both items’ dev/03 ACs flag.

response_item/reasoning also belongs here (D5, PARITY-12): its summary text, raw content chain-of-thought text when non-null (N2 — previously dropped despite this very label claiming otherwise; content is null on the vast majority of real turns, so this was easy to miss until fixtures carried the key at all), and the encrypted_content presence flag (N1: only a genuinely non-null value counts — serde_json returns Some(&Value::Null) for a present-but-null key, which is what EVERY real rollout’s reasoning item carries per upstream codex-rs/protocol/src/models.rs:970-983, so a naive .is_some() false-flagged every reasoning item as “encrypted” on real data) are captured by Session::from_codex_str onto the next assistant ChatMessage’s metadata (reasoning/ reasoning_content/reasoning_encrypted) — see audit_codex_item’s Reasoning arm. When no following assistant turn exists to attach to (a non-assistant item interrupts, or the reasoning is dangling at EOF — an aborted-turn shape, N3), it is flushed as its own synthesized [reasoning] (turn ended without a reply) message instead of being silently discarded, so this label stays honest for that shape too. It isn’t Normalized (no canonical “reasoning” ChatMessage), but it is provably not a blind drop either.

DISCLOSURE: every metadata key mentioned above (review_findings, review_overall_correctness, review_overall_confidence_score, goal_status, goal_token_budget, reasoning, reasoning_content, reasoning_encrypted) is LOADER-CAPTURE ONLY. It survives the verbatim Codex→Codex diagonal (raw bytes, untouched) and reads back out of the native supercode format, but it does NOT survive a cross-format writer or the --session-id re-serialized diagonal: ChatMessage.metadata is never serialized (message.rs:50-55) and no writer reads it back out. Don’t misread Retained here as cross-format-durable — it means “captured in-process”, not “written back out”.

Dropped

Parsed and understood, but intentionally dropped (e.g. token_count, task_started/task_complete, UI echoes of content already captured elsewhere as Normalized/Retained). See event_msg_coverage’s doc comment for the few real, currently-unrecovered exceptions (D1) — e.g. patch_apply_end’s changes[path].unified_diff — where Dropped means genuine, asserted content loss, not “duplicated elsewhere”.

Unmodeled

Not modeled at all — falls into an Unknown typed bucket.

Trait Implementations

impl Clone for Coverage

fn clone(&self) -> Coverage

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Coverage

impl Debug for Coverage

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

Formats the value using the given formatter.

impl Eq for Coverage

impl Ord for Coverage

fn cmp(&self, other: &Coverage) -> Ordering

This method returns an Ordering between self and other.

fn max(self, other: Self) -> Self

where Self: Sized,

Compares and returns the maximum of two values.

fn min(self, other: Self) -> Self

where Self: Sized,

Compares and returns the minimum of two values.

fn clamp(self, min: Self, max: Self) -> Self

where Self: Sized,

Restrict a value to a certain interval.

impl PartialEq for Coverage

fn eq(&self, other: &Coverage) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl PartialOrd for Coverage

fn partial_cmp(&self, other: &Coverage) -> Option<Ordering>

This method returns an ordering between self and other values if one exists.

fn lt(&self, other: &Rhs) -> bool

Tests less than (for self and other) and is used by the < operator.

fn le(&self, other: &Rhs) -> bool

Tests less than or equal to (for self and other) and is used by the <= operator.

fn gt(&self, other: &Rhs) -> bool

Tests greater than (for self and other) and is used by the > operator.

fn ge(&self, other: &Rhs) -> bool

Tests greater than or equal to (for self and other) and is used by the >= operator.

impl StructuralPartialEq for Coverage