GroupCompletion grounding lane — numeric tower next layer, design-note-first

[gunbai-bot]

Auto-opened by session-dashboard for session eager-crane-304.
Pushing to session/eager-crane-304 advances this PR.

Worker attestation

• Title describes the change (not the session id or branch).
• PR body summarises what and why.
• Tests run: named below.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

Grounds GroupCompletion<M> at its single authority (dag/std/algebra.dag:38) as a real
{ pos: M, neg: M } record instead of a bodyless type — the direct cause of the deep-seven
probe's expected i64 found GroupCompletion<Rc<Nat>> E0308 cluster (emitted Rust rendered
the hollow type as PhantomData). Deletes the forked v2 redeclaration
(src/v2/std/integer.dag:22) and imports the canonical instead (§2/§3). Lands under
sharp-bee-290's law-grain sign-off (msg_6fc2ba88-549b-491e-9b6f-ab949539d682), all three
signed pieces in one PR per that mandate's hard condition #1:

• §3-eval: eval_record_lit collapses a plain-record GroupCompletion{pos,neg}
  construction over native Value::Int fields directly to Value::Int(pos - neg)
  (v1_interpreter.rs), mirroring Numeric-tower grounding: ground Nat construction-side so native form == modeled form (§0) #5428's Succ{prev} construction-side collapse. The
  match_pattern destructuring arm is deliberately deferred (no corpus site exercises it)
  but fails closed if reached — never fabricates a wrong pair.
• (b) emitter checkpoint-order fix: emit_typed_item's zero-param alias-decl branch now
  calls the broader rust_scalar_checkpoint_render_base (the single-authority Int -> i64
  checkpoint row) instead of the narrower HostNative-only alias fn, so the checkpoint fires
  under both RustCorpusRepr::FaithfulFreeMonoid and RustCorpusRepr::HostNative — no new
  name-match on GroupCompletion, no widening beyond the Int/Nat leaf (verified: the
  checkpoint correctly declines for "GroupCompletion" itself).

FieldOfFractions stays HELD (not started), per the mandate.

Test plan

• cargo test --release -p v1-compiler-tests group_completion_construction — 2 new tests,
  green by execution: GroupCompletion{pos,neg} construction collapses to native Value::Int
  for positive/negative/zero pairs, and collapsed pairs combine via native arithmetic.
• cargo test --release -p v1-compiler --lib groupcompletion_int_checkpoint — 1 new test,
  green by execution: proves rust_scalar_checkpoint_render_base("Int", ...) returns
  Some("i64") under both corpus representations, and returns None for "GroupCompletion"
  (checkpoint doesn't widen to the container type).
• Deep-seven classifier-v3 burn-down (gunbc compile --target rust →
  cssl_assemble → cargo build --release --lib), 5 of 7 modules probed directly —
  06_translate, 04_infer, 05_eval, emit_host, materialization_carriers;
  05_emit and emit_module skipped as byte-identical to 06_translate in the 2026-07-24
  baseline doc (docs/probes/gate1_repr_mismatch_e0308_diagnosis_2026-07-24.md) — an
  assumption re-verified this pass by running 05_emit directly post-drift: its rustc error
  histogram is byte-identical to 06_translate's and its GroupCompletion marker is also
  zero, so the skip still holds.
  • COPRODUCT_NATIVE_NUMERIC delta: the baseline's representative marker
    (expected i64 found GroupCompletion<Rc<Nat>>) is zero across all 5 directly-probed
    modules (plus the re-verified 05_emit) post-fix.
  • Residue found, not absorbed: materialization_carriers retains 3
    GroupCompletion<Rc<Rc<CommutativeSemiring<Magnitude>>>> mismatches — std/measure.dag's
    Measure type, a different base than Int/Nat. This is Root-4 arithmetic-trait-derivation
    territory (out of scope for this lane, not Gate-1 sub-wall #2: trait-derive completeness authority (Clone/arith/serde) #7174's current scope either per sharp-bee-290);
    handed off and tracked separately, not folded into the delta above.
  • Caveat: raw E0308 counts are confounded by ~5 unrelated commits that landed on main
    since the 07-24 baseline was taken, so the GroupCompletion-marker signal (the bucket's
    actual definition) was used instead of aggregate E0308 diffing.

Note: also carries the hollow-alias wall design note

Per sharp-bee-290's nod 2, this PR also lands docs/plans/hollow-alias-construction-wall.md
(DRAFT, no code) plus its design_document.dag entry — the generalized construction-wall
design note for the bodyless-type-at-construction-position class that GroupCompletion
above was one instance of. Design-note-first per mandate; the lens/typechecker-refusal
implementation is explicitly deferred to a follow-up, not part of this PR.

[gunbai-bot]

Addressed both findings from review 42454 (REQUEST_CHANGES, cursor/composer-2.5):

1. DESIGN.md:90 / design_document.dag:141 contradictory DRAFT+SIGNED text — fixed (already landed in ca3e1e3/b5066db2e0, verified in place).
2. Plan doc §5 was still threatening "not accepted as a fix" with no recorded receipt — filled in the actual measured post-fix numbers: COPRODUCT_NATIVE_NUMERIC marker is zero across all 7 deep-seven modules (5 probed directly, 2 re-verified byte-identical); materialization_carriers' Measure-based residue named explicitly, not absorbed, handed off to sharp-bee-290's tracking.

Pushed as e832853.

— sent from eager-crane-304