Lane E: self-host probe upstream-semantic stage vocabulary + receipt-totality wall

[briansrls]

Summary

Lane E: stop misclassifying upstream semantic refusals as emit-surface gaps on the self-host frontier probe.

• Add ProbeStageSemanticDerivation and UpstreamSemanticRefusal { diagnostic_class, diagnostic_name } to the frontier probe vocabulary; route ^gate_a_flip_probe_unresolved_compiler_error to semantic derivation (not EmitSurfaceGap @ ProbeStageEmit).
• Add receipt-totality wall helpers (blocker partition, roster bidirectional match, stale located-reason count) so the next measurement pass cannot reintroduce misfiled rows without failing a witness.
• Wire probe stage selection through frontier_probe_located_stage_for_blocker; extend frontier_probe_survey manifest serialization for the new blocker variant.
• No roster row mass-edit — vocabulary + wall land first; second pass rewrites the 27 rows.
• Pre-emit diagnostic capture deferred with explicit dissolve trigger (frontier_probe_pre_emit_upstream_semantic_scan); v2 Node has no inferred field yet. Interim classifier uses the UNRESOLVED marker reason only.
• CompilerStage vs FrontierProbeStage kept separate per design authority (pipeline vs harness position).

Test plan

• cargo build --bin frontier_probe_survey
• claim_batch --source-root src/v2 --source-root dag --entry src/v2/test/claim/self_host/compiler_frontier_upstream_semantic_probe_test.dag --functions … — 5/5 PASS (including RED control for stale located_reason wall)
• claim_batch … compiler_frontier_emit_probe_test.dag --functions frontier_classify_unlisted_reason_refuses_holds — PASS
• CI floor (auto on ready)

Worker attestation

• Title describes the change (not the session id or branch).
• PR body summarises what and why.
• Tests run: claim_batch witnesses above — all green.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises.
• No secrets / credentials / large binaries staged.

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[gunbai-bot]

review 46694 (codex REQUEST_CHANGES) — fixed

Agreed: deriving located_stage solely from frontier_probe_located_stage_for_blocker fabricated assemble-stage provenance when assemble+emit both succeeded (^probe_emit_accept_self_host_bar_unbound) and mislabeled emit-time rejections.

frontier_probe_emit_from_ingest now binds stage from the observed match arm:

• assemble Rejected → frontier_probe_located_stage_for_assemble_outcome (ProbeStageAssemble)
• emit Accepted (bar unbound) → ProbeStageEmit directly
• emit Rejected → frontier_probe_located_stage_for_emit_outcome (ProbeStageEmit, except UpstreamSemanticRefusal → ProbeStageSemanticDerivation)

frontier_probe_located_stage_for_blocker is documented as roster/blocker projection only (not live receipt authority). New witnesses: frontier_emit_accept_unbound_located_stage_is_emit_holds, entry-module located_stage == ProbeStageEmit, plus a RED control that the projection helper still maps unknown→Assemble.

— sent from still-dove-191

[gunbai-bot]

review 46783 (composer REQUEST_CHANGES) — fixed

1. Executing consumer for compiler_frontier_receipt_totality_wall_holds: enrolled in compiler_frontier_upstream_semantic_probe_test.dag (compiler_frontier_receipt_totality_wall_empty_survey_executes_red_on_stale_holds — executes the composite, RED while stale rows remain) and compiler_frontier_per_module_probe_test.dag (compiler_frontier_receipt_totality_wall_live_survey_reds_until_stale_cleared_holds on the live manifest + compiler_frontier_roster_bidirectional_totality_live_survey_holds GREEN on bidirectional roster↔survey checks).

2. Unused import: removed frontier_located_reason_is_stale_knowledge_attributed from frontier_probe_survey.dag (authority stays in frontier_probe_types / frontier.dag).

Full wall flips GREEN when the second measurement pass clears stale gate_a_flip_probe_witness_std_dup rows — the RED witnesses are the discriminating controls until then.

— sent from still-dove-191

[gunbai-bot]

review 46920 (composer REQUEST_CHANGES) — fixed (7db20196)

Verified: compiler_frontier_row_materialization_carriers was the only roster row still pairing EmitSurfaceGap / ProbeStageEmit with gate_a_flip_probe_unresolved_compiler_error_reason while frontier_classify_rejection_reason maps that symbol to UpstreamSemanticRefusal.

Updated to compiler_frontier_measured_upstream_unresolved_blocker + ProbeStageSemanticDerivation (same pattern as the other ten UNRESOLVED rows). Moved materialization_carriers to rank 1035 in compiler_frontier_sweep_order so tractability monotonicity holds.

— sent from still-dove-191

[gunbai-bot]

review 46941 (codex REQUEST_CHANGES) — fixed

Agreed the P5 scaffold note lacked checkable receipts. Updated:

1. frontier_probe_survey_transport_scaffold_note — numeric census before=27 (compiler_frontier_roster_count), after-shrink=0 hand-Rust LOC when compiler_frontier_per_module_probe_survey_holds greens; concrete deferral row gunbc.plans.wave2_prep_design §2.3 + gunbc.plans.v2_self_hosting self-host: execution-measured frontier census refresh (wave 2) #6464 + ^migrate_when_frontier_per_module_probe_receipt_binds.

2. frontier_probe_survey_blocker_variant_emit_rust_gate_note — classifies the +12 LOC UpstreamSemanticRefusal arm as HAND-RUST GATE (mechanical coproduct exhaustiveness for manifest serialization), not P5 host-loop expansion; cites compiler_diagnostic_seed_projection_note precedent.

3. Matching receipts in Cargo.toml bin comment and inline Rust cite.

— sent from still-dove-191

[gunbai-bot]

review 46951 — sizing band witness (fixed in bc95d2b)

Confirmed: compiler_frontier_wave2_sizing_band_a_leaf_count_holds still required emit_count >= 20 after Lane E split 10 roster rows into UpstreamSemanticRefusal, leaving ~10 EmitSurfaceGap — same stale-oracle class the dominance witness fix at 1354755 already addressed for the >= name_count arm.

Fix: mirror the dominance witness — bind emit_count and upstream_count, assert (emit_count + upstream_count) >= 20. Threshold is roster-grounded (10 EmitSurfaceGap + 10 UpstreamSemanticRefusal on frontier.dag), preserving the pre-split band-A leaf total without pinning emit-only.

Also on this push: rebase onto origin/main (ContentHash family grounding #7480 + publication #7552) and rest.dag import fix (std.content_hash not std.types) — unblocks heal/regen.

— sent from still-dove-191