Files

T

Brummel 70f2a318e0 iter embedding-abi-m5.tidy (DONE 3/3): milestone-close doc-honesty drift — pin-safe, doc/comment-only

Resolves the M5 milestone-close audit DRIFT (audit journal
28ab56a). Single cohesive commit, M2.tidy a80d495 / M3.tidy
63d7d60 precedent (pins are the coverage, no RED, no
audit/fieldtest gate).

- docs/DESIGN.md edit-1 §"Free (host side)": dropped the retired-M4
  forward-reference ("an additive M4 concern, not a contradiction
  of this freeze"; M4 retired 2026-05-18) -> present-tense current
  fact (a boxed-field record is not an M3 embedding type,
  export-gate-rejected; the freeze covers exactly the all-scalar
  single-ctor record).
- docs/DESIGN.md edit-2 §"Embedding ABI": reconciled the now-
  inaccurate "(no shared mutable runtime state - ... data-race-free,
  sanitiser-verified)" blanket with the real post-7bfa11e state:
  the per-allocation hot path (per-ctx counters + per-object
  refcount header) is non-atomic by design and never shared
  (Ctx:!Send, single-thread-per-ctx); the one shared datum is the
  atomic-relaxed global RC-stats fallback counter; "data-race-free,
  sanitiser-verified" retained (still true).
- runtime/rc.c:88 comment-only: stale "two unconditional ++
  operations" -> accurate "one counter bump per alloc/free ...
  relaxed atomic add on the global null-ctx fallback, a plain ++
  on the per-ctx path" (consistent with the adjacent already-
  correct :93-106 atomic block + :45-55 Threading header).

Boss-verified independently: all pins green (design_schema_drift
8/0, docs_honesty_pin 5/0, effect_doc_honesty_pin 4/0,
embed_record_layout_pin 1/0) = mechanical proof no pinned/hashed
byte moved; the adjacent separately-pinned bare-scalar sentence is
byte-identical (shifted 2299->2305 by net-added lines; substring
pin, passes); rc.c strictly comment-only (filter empty);
cargo build --workspace Finished; git scope = only docs/DESIGN.md
+ runtime/rc.c + journal/stats. Clears the M5-audit doc-honesty
debt; no new debt.

FINAL M5 iteration. The M1-M5 Embedding ABI arc is functionally
complete, audited, ratified, and doc-honest. Includes the per-iter
journal, stats, and INDEX line.

2026-05-19 03:07:40 +02:00

3.6 KiB

Raw Blame History

iter embedding-abi-m5.tidy — milestone-close doc-honesty drift reconciliation

Date: 2026-05-19 Started from: e81fb90d88 Status: DONE Tasks completed: 3 of 3

Summary

The M5 milestone-close audit (2026-05-19-audit-embedding-abi-m5.md) found three doc-honesty drift items — two docs/DESIGN.md prose inaccuracies and one runtime/rc.c stale performance-aside comment. This tidy reconciles all three: DOC/COMMENT-ONLY, pin-safe, zero language/checker/codegen/schema/runtime-behaviour change. Mirrors the M2.tidy (a80d495) / M3.tidy (63d7d60) precedent — the existing pin tests (design_schema_drift, docs_honesty_pin, effect_doc_honesty_pin, embed_record_layout_pin) staying green before AND after is the mechanical proof the edits moved no pinned byte; that is the coverage, so no RED test, no audit/fieldtest gate. This is the final M5 iteration; M5 closes after Boss-verify (roadmap [~]→[x] + done-state are Boss-side).

Per-task notes

iter embedding-abi-m5.tidy.1: docs/DESIGN.md doc-honesty (edit-1
- edit-2). edit-1 §"Free (host side)" (:2356-2360): dropped the retired-M4 forward-reference ("an additive M4 concern") — M4 was retired 2026-05-18; the boxed-field record is simply export-gate-rejected (present-tense current fact). edit-2 §"Embedding ABI" (:2290-2293): reconciled the stale "no shared mutable runtime state … data-race-free, sanitiser-verified" claim with the now-atomic global RC-stats fallback — the per-allocation hot path (per-ctx counters + per-object refcount header) is non-atomic and never shared (Ctx: !Send, single-thread-per-ctx); the one shared datum is the atomic-relaxed global RC-stats fallback counter. Pin-safety proof green (8/5/4 passed, 0 failed across the three ailang-core pin suites); the adjacent separately-pinned line (Export parameters are written **bare**:, docs_honesty_pin.rs:135) is byte-identical, shifted line 2299→2305 by the net-added lines above it but content unchanged (the pin is a substring assertion, not a line-number assertion — it passes).
iter embedding-abi-m5.tidy.2: runtime/rc.c 18g.0-block (~:88-89) comment-only honesty fix. Replaced the stale "two unconditional ++ operations" performance aside with the accurate "one counter bump per alloc/free … relaxed atomic add on the global null-ctx fallback, a plain ++ on the per-ctx path", consistent with the adjacent already-correct :93-106 atomic paragraph and :45-55 Threading header. Mechanical comment-only filter clean (every changed line is a *-prefixed C comment); frozen-IR pin embed_record_layout_pin byte-identical (1 passed, 0 failed — it asserts generated IR, never rc.c comment text; M3.tidy precedent).
iter embedding-abi-m5.tidy.3: consolidated doc-only + scope gate (verification only, no files). All pins green across both crates (grep -c 'test result: ok' == 3 + embed_record_layout_pin ok); git status --porcelain shows ONLY docs/DESIGN.md + runtime/rc.c as content changes (zero crates/**, examples/**, ail-embed/**, Cargo.toml, bench/baseline.json); cargo build --workspace Finished — confirms no behaviour change.

Concerns

(none)

Known debt

(none — this iteration clears the M5-audit doc-honesty debt; no new debt introduced.)

Blocked detail

(n/a — DONE)

Files touched

docs/DESIGN.md (§"Embedding ABI" edit-2 ~:2290-2293; §"Free (host side)" edit-1 ~:2356-2360 — prose-only)
runtime/rc.c (18g.0 comment block ~:88-89 — C-comment text only)

Stats

bench/orchestrator-stats/2026-05-19-iter-embedding-abi-m5.tidy.json

3.6 KiB Raw Blame History