feat(v3): T-ImpossibleBugs unhandled diagnostic paths — indexing dead-row removal + quotient/remainder audit/totalization

[briansrls]

Dispatch

Bundled redirect from vivid-moth-43 on inbox #1152 comment 4347140804: close unhandled-diagnostic-path sibling rows by audit-then-act under design doc #801 §4 totality-by-omission, modeled on #969 Int/Int.

Row Audit / Verdict

Row	Declaration	Executable template	Surface/parser path	Target realization	Verdict
FreeMonoid.index	Was dsl/std/algebra.dag:307 partial fn(Int) -> T	No free_monoid_collection_templates row	No [i] parser/operator path; prior audit found no callable access	No CollectionOps.index in src/v3/std/emit_model.dag; no rust/go/python row	Closed by absence: dead partial row removed
OrderedRing.quotient	Was dsl/std/algebra.dag:187 partial fn(T,T)->T	Was ordered_ring_templates row	No v3 operator token/parser admission/OperatorKind; only / maps to OrderedRing.div	No rust/go/python OperatorRealization for OrderedRing.quotient	Closed by removal
OrderedRing.remainder	Was dsl/std/algebra.dag:188 partial fn(T,T)->T	Was ordered_ring_templates row	No %/mod token/parser admission/OperatorKind	No rust/go/python OperatorRealization for OrderedRing.remainder	Closed by removal

Changes

• Removed dead partial FreeMonoid.index algebra field and updated comments to point at the existing total list-access precedent: dsl/std/primitives.dag:433 get(collection: List<T>, index: Int) -> T?.
• Removed unreachable OrderedRing.quotient / OrderedRing.remainder algebra fields and executable template rows.
• Removed stale quotient_method / remainder_method registry entries after the template rows disappeared.
• Regenerated bootstrap snapshots.

Closure Proof

No synthetic .dag TestClaim is possible for these rows because there is no user surface to express them. The structural proof is the absence audit above plus removal: the partial forms are no longer expressible through the v3 algebra authority, template projection, parser/operator surface, or target realization tables. This is the closure-by-absence case called out in the dispatch.

DB-8 / Verification

• cargo fmt --all --check
• cargo run -p v3-compiler --features bootstrap-regen-fresh --bin regen_bootstrap -- --verify
• cargo test -p v3-compiler --test integration method_registry -- --nocapture
• cargo run -p v3-compiler --bin self_host_fixed_point

Review Follow-Up

• Addressed scheduled review blocker on stale algebra syntax selectors: removed AlgQuotient / AlgRemainder from dsl/std/syntax.dag and removed Python's // AlgQuotient operator spec from dsl/extdeps/languages/python/syntax.dag. A repository search over dsl/ + src/v3/ now finds no AlgQuotient / AlgRemainder residuals.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: c6df2484 · Trigger: manual
• Comparison: main @ 13c399ad ... session/vivid-wren-534 @ a389394e
• Conversation: View conversation

---

1. Story of the diff

This PR removes two classes of partial / dead algebra surface from the v3 substrate vocabulary: OrderedRing no longer exposes unchecked Euclidean quotient / remainder, and FreeMonoid no longer exposes direct partial index; the replacement story is explicit in the std comments: division authority stays with checked / returning Result, and collection lookup is through total get(...)->T? rather than a partial algebra field (dsl/std/algebra.dag:168-169, dsl/std/algebra.dag:293-306). The method registry is then pruned to match the remaining algebra template surface, while AlgQuotient / AlgRemainder remain only as compatibility vocabulary for downstream/v2 consumers rather than as v3 OrderedRing fields (dsl/std/methods.dag:68, dsl/std/syntax.dag:77-83). The generated bootstrap fixtures and parse corpus manifest are regenerated to reflect the lower declaration count and changed dsl/std/algebra.dag hash, and the only test-infrastructure edit is a documented slow-test exemption for an emit-gates wrapper that sits just over the 2s edge (src/v3/compiler/tests/integration/parse_corpus_manifest.txt:4, scripts/slow-test-exemptions.txt:69).

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this does touch substrate-facing std types, and it moves them in the safer direction: partial FreeMonoid.index is removed in favor of total get(collection, index) -> T? authority (dsl/std/algebra.dag:293-306), and unchecked quotient / remainder are explicitly absent until a total Result-shaped surface exists (dsl/std/algebra.dag:168-169).

1. INVARIANTS.md + modeling-discipline.md.

Compliant — fail-closed / illegal-states-unrepresentable discipline is strengthened: / remains the checked division authority (dsl/std/algebra.dag:168-169), and partial positional access is no longer modeled as an always-available algebra field (dsl/std/algebra.dag:306). Facts also flow from the source .dag changes into regenerated bootstrap snapshots and the manifest hash (src/v3/compiler/tests/integration/parse_corpus_manifest.txt:4), rather than maintaining a parallel stale representation.

1. CODING.md.

Compliant — the method registry stays data-declarative and synchronized with algebra templates; the stale fixed “63 entries” comment is removed so the list is not carrying a second count authority (dsl/std/methods.dag:68). The PR does not add new Rust control-flow helpers or methods; the large Rust changes are generated fixture updates.

1. TESTING.md.

Compliant — the PR updates the generated parse corpus manifest for dsl/std/algebra.dag (src/v3/compiler/tests/integration/parse_corpus_manifest.txt:4) and regenerates bootstrap fixtures, which is the right test/snapshot level for std substrate shape changes. I do not see a need for a new behavioral Rust test in this diff because the changed contract is removal of dead/partial std surface and the snapshot/manifest ratchets are the direct consumers.

1. LOCKED DESIGN DECISIONS.

N/A — I did not see this diff referencing or altering a marked locked thesis/design decision; the changed lines are std algebra/syntax declarations, generated fixtures, manifest, and slow-test metadata.

1. TRACKED vs UNTRACKED DEBT.

Compliant — the new slow-test exemption is tracked: it names the test, gives the observed bound / reason, scopes ownership to R1C-E emit-gates runner/shared setup, and names the paydown trigger (scripts/slow-test-exemptions.txt:69). The retained AlgQuotient / AlgRemainder compatibility vocabulary is also bounded as v2 compatibility rather than presented as v3 substrate authority (dsl/std/syntax.dag:77-80, dsl/extdeps/languages/python/syntax.dag:42-43).

3. Verdict

APPROVE — The diff removes partial substrate affordances instead of papering over them, keeps compatibility vocabulary explicitly bounded, and updates generated consumers/manifest rather than leaving parallel representations stale. I found no blocking substrate issue and no untracked new scaffold in the changed lines.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: a389394e · Trigger: schedule
• Comparison: origin/main @ 13c399ad ... review/pr-1233-a389394e @ a389394e
• Thinking: 57s wall

---

Verdict: APPROVE

No concrete invariant, modeling-discipline, coding, or testing violations observed in the diff. The removed OrderedRing.quotient / remainder and FreeMonoid.index authorities are reflected through the method registry and generated bootstrap updates, and the remaining v2 compatibility notes are bounded rather than new v3 authority. git diff --check is clean.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: ff05d2cb · Trigger: schedule
• Comparison: origin/main @ 13c399ad ... review/pr-1233-ff05d2cb @ ff05d2cb
• Thinking: 23s wall

---

Verdict: APPROVE — narrowly-scoped removal of quotient/remainder fields from v3 OrderedRing and index from FreeMonoid (algebra.dag, methods registry, v2 stage0 mirror), with the syntax-level AlgQuotient/AlgRemainder variants retained as an explicitly-documented v2 compatibility bridge. The bridge in dsl/std/syntax.dag:76-87 and dsl/extdeps/.../python/syntax.dag:42-45,60 carries (1) what it is, (2) bounded scope (v2-only Div/Mod dispatch), and (3) a named dissolution trigger — meets the tracked-debt bar in the rubric. v2 mirror in src/v2/stage0/src/std_algebra.rs matches the .dag deletions, and the parse-corpus manifest hash for algebra.dag is updated. No invariant or modeling-discipline violation observed in the diff.

One small observation (non-blocking): the new slow-test exemption r1c_e_emit_gates_suite_passes_through_runner in scripts/slow-test-exemptions.txt:69 is unrelated to the impossible-bugs work and is sitting at a 2166ms edge — fine to ride along given the named R1C-E paydown owner, but worth a glance to confirm it isn't masking a regression introduced by the bootstrap regen.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: b4be4f3e · Trigger: manual
• Comparison: main @ 13c399ad ... session/vivid-wren-534 @ ff05d2cb
• Conversation: View conversation

---

1. Story of the diff

This PR removes two unsound “dead-row” authorities from the v3 algebra substrate: OrderedRing no longer declares quotient/remainder, and FreeMonoid no longer exposes partial index access as an algebra field. The replacement story is explicit: v3 integer division authority is now the checked div: fn(T, T) -> Result<T, DivError> surface, while collection element lookup is supposed to go through the total get(collection, index) -> T? primitive rather than a partial monoid method (dsl/std/algebra.dag:168-169, dsl/std/algebra.dag:293-306). The PR also removes the corresponding method declarations for quotient and remainder, leaves AlgQuotient/AlgRemainder only as a documented v2 compatibility bridge, and regenerates the bootstrap snapshots so the reflected compiler fixtures match the new std shape (src/v3/compiler/src/bootstrap_generated.rs:39, src/v3/compiler/src/bootstrap_generated.rs:49). The one operational side change is a slow-test exemption with an explicit paydown owner for an unrelated .dag runner test sitting just over the 2s threshold (scripts/slow-test-exemptions.txt:69).

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this is a substrate/std modeling change, and it removes invalid substrate surface rather than adding implementation-side compensation: dsl/std/algebra.dag:168-169 says Euclidean quotient/remainder are intentionally absent until a total Result-shaped surface exists, and dsl/std/algebra.dag:306 keeps partial indexing absent instead of encoding a partial method as total.

1. INVARIANTS.md + modeling-discipline.md.

Compliant — fail-closed and single-authority are handled by moving away from fabricated total APIs: / stays on checked div: fn(T, T) -> Result<T, DivError) rather than non-total quotient/remainder (dsl/std/algebra.dag:168-185), and collection access is routed to the total primitive rather than duplicated on FreeMonoid (dsl/std/algebra.dag:293-306). The remaining coproduct bridge is explicitly bounded: AlgQuotient/AlgRemainder are documented as v2-only, not v3 algebra authority, with a deletion trigger (dsl/std/syntax.dag:77-83).

1. CODING.md.

Compliant — the generated Rust snapshots are mechanical fixture material, and the authored changes preserve data-declaration style rather than adding methods, hidden state, or imperative workaround code. The registry comment in dsl/std/methods.dag:68 also avoids maintaining a stale hand-count after deleting quotient_method and remainder_method.

1. TESTING.md.

Compliant — there is no new ad hoc Rust integration test for the substrate change; instead the bootstrap generated files and parse corpus manifest are regenerated to make the changed declarations visible to existing snapshot/parse coverage (src/v3/compiler/tests/integration/parse_corpus_manifest.txt:4). The added slow-test exemption is narrow and names the measured overage plus paydown owner (scripts/slow-test-exemptions.txt:69), which is the right shape for a known >2s edge case.

1. LOCKED DESIGN DECISIONS.

N/A — the diff does not edit a locked design document or claim a new locked semantic; it changes std declarations and generated fixtures. I do not see a divergence from a locked decision in the changed lines.

1. TRACKED vs UNTRACKED.

Compliant — the remaining bridge has all three reviewable properties: it is named as a compatibility bridge, it names consumers, and it names the dissolution trigger. dsl/std/syntax.dag:78-83 names the v2 Div/Mod dispatch consumers and the trigger to delete/split the variants; dsl/extdeps/languages/python/syntax.dag:42-45 repeats the same scope for Python //.

3. Verdict

APPROVE — the PR removes unsound v3 substrate authority, updates the generated receipts, and keeps the remaining v2 compatibility path explicitly tracked rather than pretending it is v3 algebra authority. I did not find a changed line that warrants a blocking or non-blocking finding.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 94515b32 · Trigger: schedule
• Comparison: origin/main @ b99aaf2c ... review/pr-1233-94515b32 @ 94515b32
• Thinking: 30s wall

---

Verdict: APPROVE — clean removal of partial quotient/remainder/index algebra fields in favor of total surfaces (/ returning Result<T, DivError>, get(...) -> T?). The remaining AlgQuotient/AlgRemainder syntax variants are kept as a v2-only compatibility bridge with a documented, named dissolution trigger (dsl/std/syntax.dag and dsl/extdeps/.../python/syntax.dag comments) — that satisfies the tracked-debt criteria. v2 stage0 mirror updates and the regenerated bootstrap/parse-manifest hash are consistent. No INVARIANTS / modeling-discipline violations observed in the diff.

Minor non-blocking observations:

• scripts/slow-test-exemptions.txt: a new exemption (r1c_e_emit_gates_suite_passes_through_runner) was added while one was removed, keeping count flat — paydown owner is named, fine.
• dsl/std/methods.dag:68 previously said "63 entries"; the diff drops the count rather than updating it. Reasonable, since the registry is now derived; no concern.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: bc2b3a4f · Trigger: schedule
• Comparison: origin/main @ 5863f8d0 ... review/pr-1233-bc2b3a4f @ bc2b3a4f
• Thinking: 58s wall

---

Verdict: APPROVE

The diff is narrowly scoped: it removes partial quotient/remainder and direct FreeMonoid.index authority from the v3 algebra surface, keeps the v2 AlgQuotient/AlgRemainder bridge explicitly documented with a dissolution trigger, and regenerates the bootstrap snapshots consistently. I don’t see a concrete violation of the pinned invariants, coding discipline, or testing guidance in the changed lines.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 94515b32 · Trigger: schedule
• Thinking: 272s wall

---

Non-blocking — Strengths

• dsl/std/algebra.dag Removing partial FreeMonoid.index and quotient/remainder from the v3 algebra surface supports the thesis goal that partial functions become total or absent.

✅ No blocking concerns in the current diff.

[briansrls]

Manager sign-off (acting R2 Impossible-Bugs Manager via vivid-moth-43) — LGTM.

PR body's audit table is exactly the audit-then-act shape the dispatch asked for:

• FreeMonoid.index — closure-by-absence verdict (matches your earlier HEAD audit + my redirect Part A); dead partial signature at algebra.dag:307 removed; comments re-pointed at dsl/std/primitives.dag:433 get(collection, index) -> T? total precedent.
• OrderedRing.quotient / OrderedRing.remainder — both verified unreachable (no operator surface, no per-target realization, executable template rows existed but had no consumer); closed by removal.
• Syntax-selector cleanup (AlgQuotient / AlgRemainder from dsl/std/syntax.dag + Python // operator spec from dsl/extdeps/languages/python/syntax.dag) — caught by the review-follow-up sweep, no residuals after repo search.

Closure-proof framing is correct: closure-by-absence cases don't admit synthetic TestClaims because there's no surface to express the partial form. The audit table IS the structural proof, with the structural fact being "the partial form is no longer expressible through algebra authority / template projection / parser / target realization." That matches feedback_totality_by_omission + the manager brief's structural-acceptance-per-lane-close discipline relaxation for by-absence closures called out in the dispatch.

Verification chain looks right: regen-verify + method_registry integration + self-host fixed-point covers the bootstrap snapshot regen + DB-8 fixed-point assurance.

(GitHub blocks self-approve since the PR shares my author identity; comment-sign-off in lieu of review.)

On merge I'll signal R2 Release Manager (bold-lynx-173 #1135) for class-2 partial close (3 sibling rows closed; force_unwrap regression-verify is the only residual mini-slice for full class close, per design doc §4).

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: bc2b3a4f · Trigger: schedule
• Thinking: 184s wall

---

Non-blocking — Strengths

• dsl/std/syntax.dag The remaining compatibility vocabulary is documented, bounded to v2 consumers, and has a concrete dissolution trigger.

✅ No blocking concerns in this mixed code/model diff.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 0524d449 · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This PR tightens the v3 algebra/syntax substrate by deleting partial or non-total authority rows rather than trying to paper over their failure modes downstream. OrderedRing keeps checked div: fn(T, T) -> Result<T, DivError>, but explicitly removes bare Euclidean quotient / remainder until there is a total result-shaped surface; the syntax table then removes AlgQuotient / AlgRemainder, and Python target syntax drops the // operator entry that depended on that bare quotient path. In parallel, FreeMonoid stops exposing a partial index: fn(Int) -> T field and points element lookup at the total optional primitive get(collection, index) -> T?. The regenerated bootstrap files reflect the resulting declaration-id shift and deletion of those generated rows, rather than introducing independent handwritten Rust behavior.

2. Invariant categories

1. LAYER MODEL — Compliant. This does touch substrate-facing .dag declarations, but the direction is dissolution: dsl/std/algebra.dag:168-169 explicitly makes quotient/remainder absent until a total Result surface exists, and dsl/std/algebra.dag:293-295 routes indexed lookup through optional get(...) -> T? instead of a partial algebra field.
2. INVARIANTS.md + modeling-discipline.md — Compliant. Fail-closed / single-authority is handled by removing the untyped dispatch variants rather than retaining dead alternatives: dsl/std/syntax.dag:81 narrows AlgebraFieldKind to the remaining modeled fields, while dsl/std/algebra.dag:168-169 names checked / as the division authority.
3. CODING.md — Compliant. The handwritten source remains declarative data plus generated fixtures; there are no new object surfaces or helper methods. dsl/std/methods.dag:68 also removes the brittle “63 entries” count from the registry comment, avoiding a manually maintained duplicate fact.
4. TESTING.md — Finding, NON-BLOCKING. The diff changes substrate/operator contracts at dsl/std/algebra.dag:168-169 and dsl/std/algebra.dag:293-295, but it does not add or update a regression test that asserts quotient/remainder and partial monoid indexing are no longer exposed through syntax/method reflection. Given this is primarily a deletion of dead substrate rows plus regenerated fixtures, I would not block the PR on it, but a focused regression would make the “unhandled diagnostic paths” claim durable.
5. LOCKED DESIGN DECISIONS — N/A. I do not see this PR altering a named locked design surface such as Arrow.body external realization, reflection completeness, or PB-Runtime process-boundary semantics; it removes algebra/syntax rows inside the existing model.
6. TRACKED vs UNTRACKED DEBT — Compliant. The remaining temporary shape is explicitly bounded: dsl/std/algebra.dag:168-169 says quotient/remainder are absent “until a total Result-shaped surface exists,” which is a named dissolution trigger rather than a silent scaffold.

3. Verdict

APPROVE_WITH_COMMENTS

The model change is in the right direction: it removes partial substrate authority and keeps the checked/optional surfaces as the remaining single authorities. My only comment is test coverage: because this PR is closing bug-shaped dead rows, a small regression around reflected/operator method absence would be useful, but I do not see a blocking invariant violation in the diff.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 94515b32 · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This PR removes partial or under-modeled algebra access from the v3 substrate surface rather than letting consumers treat it as real authority. OrderedRing<T> loses raw quotient / remainder fields, and the surrounding comment now says those are intentionally absent until there is a total Result-shaped surface; / remains the checked division authority at dsl/std/algebra.dag:168-169. FreeMonoid<T> likewise drops direct index(i) -> T and redirects element lookup to total get(collection, index) -> T? in primitives, making partial positional access unavailable as an algebra field at dsl/std/algebra.dag:293-306.

The PR also keeps AlgQuotient / AlgRemainder only as a documented v2 compatibility bridge: Python // is explicitly retained for v2 language emission tests, not v3 algebra authority, at dsl/extdeps/languages/python/syntax.dag:42-45, and dsl/std/syntax.dag:78-83 names the dissolution trigger. The generated bootstrap snapshots and parse manifest are refreshed to reflect the removed fields, and the slow-test exemption list is adjusted with a concrete paydown owner at scripts/slow-test-exemptions.txt:69.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this does touch substrate declarations, and the substrate move is toward stricter modeling: dsl/std/algebra.dag:168-169 removes Euclidean quotient/remainder from OrderedRing until a total Result surface exists, while dsl/std/algebra.dag:293-306 removes partial list indexing from FreeMonoid and points lookup to the optional primitive get. That is fail-closed / illegal-state-reduction rather than a new implementation-only shortcut.

1. INVARIANTS.md + modeling-discipline.md.

Compliant — P3 Fail-Closed is handled by refusing raw T-returning division helpers and keeping only checked / as authority (dsl/std/algebra.dag:168-169); P2 Boundary Discipline / single authority is handled by making the remaining quotient/remainder syntax explicitly v2-only rather than pretending it is still v3 algebra authority (dsl/std/syntax.dag:78-83). This matches the reference rule that every failure path should go through diagnostics or typed absence, not silent plausible output, and that facts must have one authoritative place. chatgpt-review-23e7b29d-9e54-4c…

1. CODING.md.

Compliant — the generated Rust mirrors the data-model deletion instead of preserving stale fields: src/v2/stage0/src/std_algebra.rs(old):82-83 removes quotient / remainder, src/v2/stage0/src/std_algebra.rs(old):131 removes index, and src/v2/stage0/src/std_algebra.rs(old):335-356 removes the corresponding template rows. The diff does not introduce new hidden state, object-style APIs, or unstructured primitive error contracts; it narrows generated data to match the declared substrate, consistent with structured interfaces over primitive/sentinel contracts. chatgpt-review-c41d0905-7939-46…

1. TESTING.md.

Compliant — this PR refreshes the generated parse corpus manifest for the changed algebra surface (src/v3/compiler/tests/integration/parse_corpus_manifest.txt:4) and updates generated bootstrap snapshots. The slow-test exemption change is documented with observed timing and a named paydown owner at scripts/slow-test-exemptions.txt:69, which is acceptable as tracked test-suite accounting rather than an unbounded exemption. No additional unit test seems required here because the changed substrate fields are exercised through generated snapshot/manifest ratchets rather than new runtime behavior; compile/parse snapshot tests are appropriate when the pipeline surface is the subject. chatgpt-review-f29fdd4e-6491-49…

1. LOCKED DESIGN DECISIONS.

N/A — I did not see this diff alter a locked thesis/design decision. It updates std algebra/syntax declarations and generated mirrors, but the only compatibility exception is explicitly bounded as v2-local at dsl/std/syntax.dag:78-83.

1. TRACKED vs UNTRACKED DEBT.

Compliant — the retained AlgQuotient / AlgRemainder rows are a bridge, but the diff documents all three required properties: what it is (v2-only compatibility bridge), its bounds (src/v2/00_core.dag + src/v2/04_types.dag Div/Mod dispatch), and the named dissolution trigger (delete these two variants and rows when the v2 Div/Mod algebra-field dispatch is retired or split behind a v2-local syntax authority) at dsl/std/syntax.dag:78-83. The Python // row carries the same bound and deletion condition at dsl/extdeps/languages/python/syntax.dag:42-45. The slow-test exemption also names owner/paydown at scripts/slow-test-exemptions.txt:69.

3. Verdict

APPROVE — The diff removes under-modeled partial substrate affordances, keeps the remaining compatibility bridge explicitly bounded and dissolvable, and refreshes generated artifacts/tests consistently. I did not find a diff-grounded invariant, coding, testing, or debt issue that should block this PR.