Repository navigation
Body-emit Track B next slice: continue #6840 (general body producer, Stage B cross-decl) toward the 12 emit-shape-gated tail modules; re-probe after the slice converts refusals to flips - #6859
Merged
Conversation
…projections. Reorder body lowering so fold-family calls desugar to seam Loop before generic Transform lowering; extend the accumulator-copy analyze walk with lowered Loop/ Transform arms plus surface postfix projection classification (helper.step). Co-authored-by: Cursor <cursoragent@cursor.com>
gunbai-bot
Bot
force-pushed
the
session/wise-bee-713-stage-b-projection-lens
branch
from
July 18, 2026 17:34
242e10f to
cc81fdc
Compare
…very budget). self_host/ per-PR discovery enforces a 5s eval cap; tokenize→parse→normalize projection receipts run ~10s and belong in long/ like peer fn_body seam witnesses. Co-authored-by: Cursor <cursoragent@cursor.com>
Contributor
|
Review 39663 (cursor/composer-2.5 APPROVE) — verified on current HEAD Re-checked each claim against the tree; no code changes needed for this review.
Post-review commit ( Awaiting CI green on run 29654443333. — sent from wise-bee-713 |
Drop unused tree_contains_surface_postfix_expr and tree_contains_lowered_transform; the latter's test import was never referenced. Co-authored-by: Cursor <cursoragent@cursor.com>
Contributor
|
Reviews 39665 + 39666 — addressed on
— sent from wise-bee-713 |
Contributor
|
Review 39669 (composer-2.5, APPROVE) — verified on current HEAD Re-checked each claim against the tree; no code changes needed.
No merge-blocking issues; dead-code cleanup from review 39666 already landed in the same HEAD. — sent from wise-bee-713 |
Contributor
|
Review 39670 (claude-opus-4-7, APPROVE) — verified on current HEAD Re-checked each claim against the tree; no code changes needed.
CI run 29655014078: — sent from wise-bee-713 |
briansrls
pushed a commit
that referenced
this pull request
Jul 19, 2026
…to std_dup bucket Re-probed all seven Gate-A candidates with CSSL_STD_SEED_LINK=1 on integration/sharp-bee-290. No cargo-green flips (baseline stays 6). #6868 Ord fix cleared btree_set_ord for 04_infer/program_partition; both now refuse on Witness std_dup alongside 01_tokenize. Four modules still refuse UNRESOLVED_CompilerError. Bank execution receipts in docs/probes/gate_a_reprobe_2026-07-19.tsv and tail_reprobe_2026-07-19.tsv. Co-authored-by: Cursor <cursoragent@cursor.com>
4 tasks done
briansrls
pushed a commit
that referenced
this pull request
Jul 19, 2026
briansrls
pushed a commit
that referenced
this pull request
Jul 19, 2026
briansrls
added a commit
that referenced
this pull request
Jul 19, 2026
…emitter Ord unlock (single merge point) (#6866) * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * emit: generalize BTreeSet Ord eligibility for Symbol-wrapped carriers FormalNonterminal and FormalTerminal ({ identity: Symbol }) now pass rust_btree_set_element_ord_eligible and receive Ord derives, unblocking 04_infer and program_partition self-emit cargo probes. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Module identity Phase 0(b): the admission invariant — every witness row * Module identity Phase 1: the path⇄module binding authority (parse-derived, fail-closed) Homes the path⇄module binding on v2.compiler.source_authority as a DERIVED-at-parse fact per docs/plans/module-identity-storage-binding-design.md §2. - ModuleStorageProvenance = ParsedFromSource { artifact, span_index } | ProducedByBehavior { producer } — zero-file produced modules are now representable honestly instead of being excluded. - ModuleStorageBinding / ModuleStorageIndex, reusing QualifiedName, Artifact, SourceRootRef and SpanIndex (no new nickname, §3). - Both projections: file -> module and module -> storage. - Scope is the LIVE 1:1 binding; many-to-many is a named deferral (§6). §5 fail-closed repair in v2.compiler.program_assembly: the Rejected arm of qualified_name_from_module_node kept the root and left the index UNCHANGED, so a module whose name could not be derived went silently invisible to every downstream lookup — an absorbing fallback in drop form, its frequency zeroed by construction. It now refuses with the located diagnostic. Witnesses (green by execution, REDs verified discriminating by perturbation): derivation from parse; path projection; duplicate-module refuses; underivable name refuses; produced module has no storage. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Consolidate the qualified_name_from_segment_list §3 fork (LIVE runtime h * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 0(b): the admission invariant — every witness row * review: forward located diagnostics; hoist grammar prepare; split fast/long witness lanes Three fixes from review + CI. 1. Located refusal (cursor/composer-2.5). The name-derivation Rejected arm discarded the derivation's diagnostics and substituted a port-level source_authority_diagnostic — satisfying "typed" while destroying "located" (§5 requires both), contradicting this PR's own notes, and diverging from the sibling arm in program_assembly which forwards d. It now forwards d unchanged, so one failure yields one diagnostic regardless of entry point. 2. Grammar prepare hoisted above the fold, mirroring program_assembly_prepare_once_note. The derivation called parse_module per read, re-validating the grammar K times over a K-module ingest; it now prepares once and uses parse_module_prepared. Empty ingest is guarded so it never pays the prepare for zero reads. Same cost-shape defect the assembly fold already documents (§6 always-fix). 3. Witness lanes split per the operator 5-second rule (CI EvalBudgetExceeded: 4 witnesses at ~5002ms). Measured by differencing against the non-parsing witness, parse-derived eval is ~13s — inherently over the fast-lane budget even after fix 2. Per gunbc.ci_layer_roots long_lane_exclusion_note ("fast receipts stay discovered, execution proofs live in long/"), the parse-derived proofs move to src/v2/test/claim/long/, and the fast lane keeps NEW witnesses that exercise the §5 construction wall over hand-built rows at zero parse cost: duplicate-module refusal, its accept control, and the produced/zero-file arm. Duplicate detection keys on module identity independently of provenance, which is why the wall is provable without a parse. All eight witnesses green by execution across both lanes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * Gate-A re-probe post-#6859+#6868: repoint 04_infer/program_partition to std_dup bucket Re-probed all seven Gate-A candidates with CSSL_STD_SEED_LINK=1 on integration/sharp-bee-290. No cargo-green flips (baseline stays 6). #6868 Ord fix cleared btree_set_ord for 04_infer/program_partition; both now refuse on Witness std_dup alongside 01_tokenize. Four modules still refuse UNRESOLVED_CompilerError. Bank execution receipts in docs/probes/gate_a_reprobe_2026-07-19.tsv and tail_reprobe_2026-07-19.tsv. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): add cssl_std_seed_link and shim_lib_rel columns to re-probe TSVs Invocation metadata for std_dup fix lane: all 19 probes were raw-closure (CSSL_STD_SEED_LINK=1, shim_lib_rel empty — no lane shims passed). Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * integration: address review 39722/39727 — admission scanner refuses (never widens), provenance de-fork, stale note, KeySource import Review 39722 (claude, RC) three findings + review 39727 (cursor, RC) two findings, all verified: 1. §5 absorbing fallback in the witness-admission scanner: rewritten as per-form structural extraction that is fail-closed BOTH ways — every occurrence of a recognized row head (bin_wet, probe_red, self_host_wet_entry, SelfHostWetReceiptBinding) either parses to a key, is a verified definition/non-literal pass-through site, or PANICS with source label + byte offset; the catch-all entry/function loop (the widen arm that could silently excuse orphans) is DELETED. A consumer in an unrecognized form now surfaces as a loud orphan, never an absorbed excuse. Call sites carry source labels (the 39727 arity finding was the auto-WIP intermediate state; complete here). 2. §3 parallel authority (borderline per reviewer): the dissolve-on marker now names the tracked lane (module-binding supply-carrier pattern; host consumes emitted manifest rows) and the interim scan's fail-closed semantics. 3. §2 byte-identical provenance pair: source_ir_node_artifact_provenance_add_from_model + _add wrapper DELETED (pre-existing on main, not introduced by the wave — verified via git show origin/main); consumers repointed (edit_locus_resolver_test import + 2 calls, internal :627 site). Zero references remain. 4. Stale fork note in module_storage_binding_derivation_test updated: the qualified_name fork it described was dissolved in this wave (renamed apart); note now records the resolution. 5. Operator-requested: dag/std/realization.dag KeySource unlisted-import x3 (pre-existing main defect) fixed — KeySource added to the std.effects import list; source_authority closure compile-clean 3 -> 0 diagnostics. Receipts: cargo build release clean; admission unit tests 2/2 (witness_admission_deferred_rows_have_consumers, witness_admission_orphan_synthetic_row_refuses); module_storage_binding witnesses PASS; edit_locus_resolver witnesses PASS on the repointed name. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * integration: rustfmt the admission scanner Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: v1-deletion dependency graph as .dag — milestones, serial-vs-fanout, critical path (operator 2-week target) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: record 2026-07-19 census discharges — Ord fix cleared btree_set_ord, KeySource import fix cleared the whole unresolved bucket; ~19 modules now on the one std_dup fix Verified by execution at f071536: all 6 formerly-UNRESOLVED closures (00_compile, materialization_carriers, program_assembly, source_authority, 02_parse, 03_ingest) compile 0 diagnostics; re-probe TSVs (#6872) show 04_infer/program_partition emit clean with std_dup as sole refusal. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Brian Searls <briansrls@gunb.ai> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
briansrls
added a commit
that referenced
this pull request
Jul 19, 2026
…n-note disposition (single merge point) (#6885) * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * emit: generalize BTreeSet Ord eligibility for Symbol-wrapped carriers FormalNonterminal and FormalTerminal ({ identity: Symbol }) now pass rust_btree_set_element_ord_eligible and receive Ord derives, unblocking 04_infer and program_partition self-emit cargo probes. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Module identity Phase 0(b): the admission invariant — every witness row * Module identity Phase 1: the path⇄module binding authority (parse-derived, fail-closed) Homes the path⇄module binding on v2.compiler.source_authority as a DERIVED-at-parse fact per docs/plans/module-identity-storage-binding-design.md §2. - ModuleStorageProvenance = ParsedFromSource { artifact, span_index } | ProducedByBehavior { producer } — zero-file produced modules are now representable honestly instead of being excluded. - ModuleStorageBinding / ModuleStorageIndex, reusing QualifiedName, Artifact, SourceRootRef and SpanIndex (no new nickname, §3). - Both projections: file -> module and module -> storage. - Scope is the LIVE 1:1 binding; many-to-many is a named deferral (§6). §5 fail-closed repair in v2.compiler.program_assembly: the Rejected arm of qualified_name_from_module_node kept the root and left the index UNCHANGED, so a module whose name could not be derived went silently invisible to every downstream lookup — an absorbing fallback in drop form, its frequency zeroed by construction. It now refuses with the located diagnostic. Witnesses (green by execution, REDs verified discriminating by perturbation): derivation from parse; path projection; duplicate-module refuses; underivable name refuses; produced module has no storage. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Consolidate the qualified_name_from_segment_list §3 fork (LIVE runtime h * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 0(b): the admission invariant — every witness row * review: forward located diagnostics; hoist grammar prepare; split fast/long witness lanes Three fixes from review + CI. 1. Located refusal (cursor/composer-2.5). The name-derivation Rejected arm discarded the derivation's diagnostics and substituted a port-level source_authority_diagnostic — satisfying "typed" while destroying "located" (§5 requires both), contradicting this PR's own notes, and diverging from the sibling arm in program_assembly which forwards d. It now forwards d unchanged, so one failure yields one diagnostic regardless of entry point. 2. Grammar prepare hoisted above the fold, mirroring program_assembly_prepare_once_note. The derivation called parse_module per read, re-validating the grammar K times over a K-module ingest; it now prepares once and uses parse_module_prepared. Empty ingest is guarded so it never pays the prepare for zero reads. Same cost-shape defect the assembly fold already documents (§6 always-fix). 3. Witness lanes split per the operator 5-second rule (CI EvalBudgetExceeded: 4 witnesses at ~5002ms). Measured by differencing against the non-parsing witness, parse-derived eval is ~13s — inherently over the fast-lane budget even after fix 2. Per gunbc.ci_layer_roots long_lane_exclusion_note ("fast receipts stay discovered, execution proofs live in long/"), the parse-derived proofs move to src/v2/test/claim/long/, and the fast lane keeps NEW witnesses that exercise the §5 construction wall over hand-built rows at zero parse cost: duplicate-module refusal, its accept control, and the produced/zero-file arm. Duplicate detection keys on module identity independently of provenance, which is why the wall is provable without a parse. All eight witnesses green by execution across both lanes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * Gate-A re-probe post-#6859+#6868: repoint 04_infer/program_partition to std_dup bucket Re-probed all seven Gate-A candidates with CSSL_STD_SEED_LINK=1 on integration/sharp-bee-290. No cargo-green flips (baseline stays 6). #6868 Ord fix cleared btree_set_ord for 04_infer/program_partition; both now refuse on Witness std_dup alongside 01_tokenize. Four modules still refuse UNRESOLVED_CompilerError. Bank execution receipts in docs/probes/gate_a_reprobe_2026-07-19.tsv and tail_reprobe_2026-07-19.tsv. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): add cssl_std_seed_link and shim_lib_rel columns to re-probe TSVs Invocation metadata for std_dup fix lane: all 19 probes were raw-closure (CSSL_STD_SEED_LINK=1, shim_lib_rel empty — no lane shims passed). Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * integration: address review 39722/39727 — admission scanner refuses (never widens), provenance de-fork, stale note, KeySource import Review 39722 (claude, RC) three findings + review 39727 (cursor, RC) two findings, all verified: 1. §5 absorbing fallback in the witness-admission scanner: rewritten as per-form structural extraction that is fail-closed BOTH ways — every occurrence of a recognized row head (bin_wet, probe_red, self_host_wet_entry, SelfHostWetReceiptBinding) either parses to a key, is a verified definition/non-literal pass-through site, or PANICS with source label + byte offset; the catch-all entry/function loop (the widen arm that could silently excuse orphans) is DELETED. A consumer in an unrecognized form now surfaces as a loud orphan, never an absorbed excuse. Call sites carry source labels (the 39727 arity finding was the auto-WIP intermediate state; complete here). 2. §3 parallel authority (borderline per reviewer): the dissolve-on marker now names the tracked lane (module-binding supply-carrier pattern; host consumes emitted manifest rows) and the interim scan's fail-closed semantics. 3. §2 byte-identical provenance pair: source_ir_node_artifact_provenance_add_from_model + _add wrapper DELETED (pre-existing on main, not introduced by the wave — verified via git show origin/main); consumers repointed (edit_locus_resolver_test import + 2 calls, internal :627 site). Zero references remain. 4. Stale fork note in module_storage_binding_derivation_test updated: the qualified_name fork it described was dissolved in this wave (renamed apart); note now records the resolution. 5. Operator-requested: dag/std/realization.dag KeySource unlisted-import x3 (pre-existing main defect) fixed — KeySource added to the std.effects import list; source_authority closure compile-clean 3 -> 0 diagnostics. Receipts: cargo build release clean; admission unit tests 2/2 (witness_admission_deferred_rows_have_consumers, witness_admission_orphan_synthetic_row_refuses); module_storage_binding witnesses PASS; edit_locus_resolver witnesses PASS on the repointed name. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * integration: rustfmt the admission scanner Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: v1-deletion dependency graph as .dag — milestones, serial-vs-fanout, critical path (operator 2-week target) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: record 2026-07-19 census discharges — Ord fix cleared btree_set_ord, KeySource import fix cleared the whole unresolved bucket; ~19 modules now on the one std_dup fix Verified by execution at f071536: all 6 formerly-UNRESOLVED closures (00_compile, materialization_carriers, program_assembly, source_authority, 02_parse, 03_ingest) compile 0 diagnostics; re-probe TSVs (#6872) show 04_infer/program_partition emit clean with std_dup as sole refusal. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(cssl): assembly std_dup unlock — Witness sanitize + v2_std emit-retain Single assembly fix for the 13-module std_dup gate (calm-boar-697 census): - sanitize_witness_import_conflict matches `pub enum Witness<C>` (generic), stripping v1_rt::Witness prelude imports that collided with emitted enum - v2_std_* emit-retain instead of minimal bridge shims (broken dependents like v2_std_logic → v2_std_algebra re-exports) No emitter changes. Probes move past HARNESS_ARTIFACT_std_dup / E0255 on all Gate-A modules + tail representatives; next blockers surface as namespace gaps. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): post-std_dup re-probe on proud-deer ed68594 (13 modules) CSSL_STD_SEED_LINK=1, shim_lib_rel empty, rebuilt cssl_assemble. std_dup E0255 cleared on 12/13; new namespace layer surfaces (FreeMonoid, ResolvedTree, InferredTree/InferredFacts). emit_produced: Optional std_dup residue. Harness invocation contract in probe script header. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): baseline caveat + probe_notes on post-std_dup TSV (#6884) ed68594 predates #6883 emit-retain; mark ResolvedTree/InferredTree rows EXPECTED-CLEARED. emit_produced Optional routed to proud-deer. FreeMonoid rows flagged as live next-layer census post-#6883. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(#6880): cargo-stage unresolved re-probe + plan census correction on main Rebased onto main post-#6866. Remaining delta: KeySource cargo-stage TSV, embedded-marker grep receipt, v1_deletion_plan corrections (13 std_dup vs 6 unresolved, ROOT-CAUSED emitter-gap, stern-lark-430 dispatched). Gate-A/tail frontier rows already landed via integration merge. Co-authored-by: Cursor <cursoragent@cursor.com> * witness_admission: bounded dissolution note on the two single-arm predicates (review 39758 disposition) Phase 0(b) author closed out; disposition owned by sharp-bee-290 per the capture. Pattern matches target_value_expr_int_literal_predicate_dissolution_note. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan(#6880): resolve integration merge — std_dup LANDED #6876, flip-wave unblocked Fold integration/sharp-bee-290 at 81dc2d9; update v1_deletion_plan lane_state to current truth (13+6 census retained, probe_flip_fanout unblocked pending emitter fixes). Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Brian Searls <briansrls@gunb.ai> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
briansrls
added a commit
that referenced
this pull request
Jul 19, 2026
#6880) * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Module identity Phase 0(b): the admission invariant — every witness row * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * emit: generalize BTreeSet Ord eligibility for Symbol-wrapped carriers FormalNonterminal and FormalTerminal ({ identity: Symbol }) now pass rust_btree_set_element_ord_eligible and receive Ord derives, unblocking 04_infer and program_partition self-emit cargo probes. Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Module identity Phase 0(b): the admission invariant — every witness row * Module identity Phase 1: the path⇄module binding authority (parse-derived, fail-closed) Homes the path⇄module binding on v2.compiler.source_authority as a DERIVED-at-parse fact per docs/plans/module-identity-storage-binding-design.md §2. - ModuleStorageProvenance = ParsedFromSource { artifact, span_index } | ProducedByBehavior { producer } — zero-file produced modules are now representable honestly instead of being excluded. - ModuleStorageBinding / ModuleStorageIndex, reusing QualifiedName, Artifact, SourceRootRef and SpanIndex (no new nickname, §3). - Both projections: file -> module and module -> storage. - Scope is the LIVE 1:1 binding; many-to-many is a named deferral (§6). §5 fail-closed repair in v2.compiler.program_assembly: the Rejected arm of qualified_name_from_module_node kept the root and left the index UNCHANGED, so a module whose name could not be derived went silently invisible to every downstream lookup — an absorbing fallback in drop form, its frequency zeroed by construction. It now refuses with the located diagnostic. Witnesses (green by execution, REDs verified discriminating by perturbation): derivation from parse; path projection; duplicate-module refuses; underivable name refuses; produced module has no storage. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Consolidate the qualified_name_from_segment_list §3 fork (LIVE runtime h * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 1: the path⇄module binding authority — parse-deriv * WIP: Module identity Phase 0(b): the admission invariant — every witness row * review: forward located diagnostics; hoist grammar prepare; split fast/long witness lanes Three fixes from review + CI. 1. Located refusal (cursor/composer-2.5). The name-derivation Rejected arm discarded the derivation's diagnostics and substituted a port-level source_authority_diagnostic — satisfying "typed" while destroying "located" (§5 requires both), contradicting this PR's own notes, and diverging from the sibling arm in program_assembly which forwards d. It now forwards d unchanged, so one failure yields one diagnostic regardless of entry point. 2. Grammar prepare hoisted above the fold, mirroring program_assembly_prepare_once_note. The derivation called parse_module per read, re-validating the grammar K times over a K-module ingest; it now prepares once and uses parse_module_prepared. Empty ingest is guarded so it never pays the prepare for zero reads. Same cost-shape defect the assembly fold already documents (§6 always-fix). 3. Witness lanes split per the operator 5-second rule (CI EvalBudgetExceeded: 4 witnesses at ~5002ms). Measured by differencing against the non-parsing witness, parse-derived eval is ~13s — inherently over the fast-lane budget even after fix 2. Per gunbc.ci_layer_roots long_lane_exclusion_note ("fast receipts stay discovered, execution proofs live in long/"), the parse-derived proofs move to src/v2/test/claim/long/, and the fast lane keeps NEW witnesses that exercise the §5 construction wall over hand-built rows at zero parse cost: duplicate-module refusal, its accept control, and the produced/zero-file arm. Duplicate detection keys on module identity independently of provenance, which is why the wall is provable without a parse. All eight witnesses green by execution across both lanes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * WIP: Emitter Ord-eligibility unlock: rust_btree_set_element_ord_eligible cons * Gate-A re-probe post-#6859+#6868: repoint 04_infer/program_partition to std_dup bucket Re-probed all seven Gate-A candidates with CSSL_STD_SEED_LINK=1 on integration/sharp-bee-290. No cargo-green flips (baseline stays 6). #6868 Ord fix cleared btree_set_ord for 04_infer/program_partition; both now refuse on Witness std_dup alongside 01_tokenize. Four modules still refuse UNRESOLVED_CompilerError. Bank execution receipts in docs/probes/gate_a_reprobe_2026-07-19.tsv and tail_reprobe_2026-07-19.tsv. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): add cssl_std_seed_link and shim_lib_rel columns to re-probe TSVs Invocation metadata for std_dup fix lane: all 19 probes were raw-closure (CSSL_STD_SEED_LINK=1, shim_lib_rel empty — no lane shims passed). Co-authored-by: Cursor <cursoragent@cursor.com> * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * WIP: Weak Self Host -> Strong Self Host (Wave 1 -> 4, on 1 exit) * integration: address review 39722/39727 — admission scanner refuses (never widens), provenance de-fork, stale note, KeySource import Review 39722 (claude, RC) three findings + review 39727 (cursor, RC) two findings, all verified: 1. §5 absorbing fallback in the witness-admission scanner: rewritten as per-form structural extraction that is fail-closed BOTH ways — every occurrence of a recognized row head (bin_wet, probe_red, self_host_wet_entry, SelfHostWetReceiptBinding) either parses to a key, is a verified definition/non-literal pass-through site, or PANICS with source label + byte offset; the catch-all entry/function loop (the widen arm that could silently excuse orphans) is DELETED. A consumer in an unrecognized form now surfaces as a loud orphan, never an absorbed excuse. Call sites carry source labels (the 39727 arity finding was the auto-WIP intermediate state; complete here). 2. §3 parallel authority (borderline per reviewer): the dissolve-on marker now names the tracked lane (module-binding supply-carrier pattern; host consumes emitted manifest rows) and the interim scan's fail-closed semantics. 3. §2 byte-identical provenance pair: source_ir_node_artifact_provenance_add_from_model + _add wrapper DELETED (pre-existing on main, not introduced by the wave — verified via git show origin/main); consumers repointed (edit_locus_resolver_test import + 2 calls, internal :627 site). Zero references remain. 4. Stale fork note in module_storage_binding_derivation_test updated: the qualified_name fork it described was dissolved in this wave (renamed apart); note now records the resolution. 5. Operator-requested: dag/std/realization.dag KeySource unlisted-import x3 (pre-existing main defect) fixed — KeySource added to the std.effects import list; source_authority closure compile-clean 3 -> 0 diagnostics. Receipts: cargo build release clean; admission unit tests 2/2 (witness_admission_deferred_rows_have_consumers, witness_admission_orphan_synthetic_row_refuses); module_storage_binding witnesses PASS; edit_locus_resolver witnesses PASS on the repointed name. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * integration: rustfmt the admission scanner Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: v1-deletion dependency graph as .dag — milestones, serial-vs-fanout, critical path (operator 2-week target) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan: record 2026-07-19 census discharges — Ord fix cleared btree_set_ord, KeySource import fix cleared the whole unresolved bucket; ~19 modules now on the one std_dup fix Verified by execution at f071536: all 6 formerly-UNRESOLVED closures (00_compile, materialization_carriers, program_assembly, source_authority, 02_parse, 03_ingest) compile 0 diagnostics; re-probe TSVs (#6872) show 04_infer/program_partition emit clean with std_dup as sole refusal. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(cssl): assembly std_dup unlock — Witness sanitize + v2_std emit-retain Single assembly fix for the 13-module std_dup gate (calm-boar-697 census): - sanitize_witness_import_conflict matches `pub enum Witness<C>` (generic), stripping v1_rt::Witness prelude imports that collided with emitted enum - v2_std_* emit-retain instead of minimal bridge shims (broken dependents like v2_std_logic → v2_std_algebra re-exports) No emitter changes. Probes move past HARNESS_ARTIFACT_std_dup / E0255 on all Gate-A modules + tail representatives; next blockers surface as namespace gaps. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): post-std_dup re-probe on proud-deer ed68594 (13 modules) CSSL_STD_SEED_LINK=1, shim_lib_rel empty, rebuilt cssl_assemble. std_dup E0255 cleared on 12/13; new namespace layer surfaces (FreeMonoid, ResolvedTree, InferredTree/InferredFacts). emit_produced: Optional std_dup residue. Harness invocation contract in probe script header. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(probes): baseline caveat + probe_notes on post-std_dup TSV (#6884) ed68594 predates #6883 emit-retain; mark ResolvedTree/InferredTree rows EXPECTED-CLEARED. emit_produced Optional routed to proud-deer. FreeMonoid rows flagged as live next-layer census post-#6883. Co-authored-by: Cursor <cursoragent@cursor.com> * docs(#6880): cargo-stage unresolved re-probe + plan census correction on main Rebased onto main post-#6866. Remaining delta: KeySource cargo-stage TSV, embedded-marker grep receipt, v1_deletion_plan corrections (13 std_dup vs 6 unresolved, ROOT-CAUSED emitter-gap, stern-lark-430 dispatched). Gate-A/tail frontier rows already landed via integration merge. Co-authored-by: Cursor <cursoragent@cursor.com> * witness_admission: bounded dissolution note on the two single-arm predicates (review 39758 disposition) Phase 0(b) author closed out; disposition owned by sharp-bee-290 per the capture. Pattern matches target_value_expr_int_literal_predicate_dissolution_note. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * plan(#6880): resolve integration merge — std_dup LANDED #6876, flip-wave unblocked Fold integration/sharp-bee-290 at 81dc2d9; update v1_deletion_plan lane_state to current truth (13+6 census retained, probe_flip_fanout unblocked pending emitter fixes). Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Brian Searls <briansearls1@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Stage B cross-decl for #6840 / Track B: fold-family calls desugar to seam
Loopbefore genericTransformlowering (body_lowering_fold.dag), and the accumulator-copy lens walks loweredLoop/Transformsubstrate plus surface postfix projection calls (helper.step) so carrier classification is fail-closed inside ingested fold bodies (projection_call_carrier_is_classified).Test plan
claim_batch --source-root src/v2 --source-root dag --entry src/v2/test/claim/long/accumulator_copy_fold_analysis_test.dag— projection + fold regression witnesses PASS locally8c1cfdbe9)Long-lane witness completion receipt (
8c1cfdbe9)Moved
wave1_gate1_a1_projection_call_witness_test.dagfromself_host/→long/because tokenize→parse→normalize exceeds the 5s per-PR discovery cap (~10s). Re-laning requires a green completion run, not just a timeout:Scope wall: no emitter (
05_emit_rust.dag) changes in this PR.