Resolves the M2 milestone-close audit's only actionable drift item
(architect [medium] DESIGN.md:2266 stale (M1) header vs its now-M2
present-tense body + coupled [low] :2354 xref). Current-state-mirror:
the section describes the current embedding ABI (M1+M2), so the
milestone tag is internal history, not current state.
5 live-coupling sites renamed in lockstep, 3 files:
- docs/DESIGN.md:2266 header '## Embedding ABI (M1)' -> '## Embedding ABI'
- docs/DESIGN.md:2354 schema-block xref 'see §"Embedding ABI"'
- docs_honesty_pin.rs:134 test comment + :136 assert-message
- crates/ailang-core/specs/form_a.md:89 live forward-xref (Boss-added
5th site: plan-recon-undercount countermeasure surfaced a live
dangling-xref the architect's 3-site scope missed; folded in on
the merits — a half-rename contradicts the tidy's own thesis)
docs_honesty_pin.rs:135 asserted substring (the rename-immune
DESIGN.md body sentence) byte-verbatim untouched. Committed
append-only history NOT falsified (Boss-verified diff = 3 content
files + journal + stats only). Pins green before AND after
(docs_honesty_pin 5/0, design_schema_drift 8/0). Zero
language/checker/codegen/schema change.
Architect drift_found: the only actionable item is the stale
'## Embedding ABI (M1)' DESIGN.md header (:2266) vs its now-M2 body
+ coupled :2354 xref — a current-state-mirror honesty desync the
iter journal flagged truthfully (Task-7-scope-deferred to preserve
the docs_honesty_pin:135 verbatim sentence). Everything substantive
holds (Invariant 1, internal-convention byte-invariance, Decision-10
note present-tense-correct, lockstep, no unrequested extras).
Bench check.py exit 1 (bump_s + 2x rc-latency max_us). M2 changed
the benched runtime path so byte-identical exoneration did not apply
-> ailang-bencher localised: H0 refuted, all 3 firings are the two
tracked roadmap-P2 noise families (bump_s causally impossible from
M2's diff; rc.c hunk branchless+free; max_us median/p99 flat +
non-reproducible). NO baseline ratified; not M2 artefacts.
Resolution: doc-honesty drift -> tidy iter embedding-abi-m2.tidy;
bench -> carry-on. Roadmap [~]->[x] after the tidy lands clean.
Completes Embedding ABI — M2.1 (Tasks 1-4 committed c9a84b3 as the
known-good subset; this commit lands Tasks 5-9 from the re-dispatch
against the Boss-corrected plan).
- T5: swarm.c per-thread-ctx scalar-swarm capability demo (no
SHARED_CTX) + rc_accounting_shared_ctx_is_flagged_by_tsan added to
the item-1 harness driver (de-globalisation negative-control teeth,
where a shared ctx genuinely races on ctx->alloc_count).
- T6: M1 host.c migrated to the ctx ABI; embed_e2e green (s==25
through the ctx-threaded forwarder).
- T7: DESIGN.md current-state update (provisional-until-M3 narrowed
to the value/record layout; Decision-10 per-thread-ctx note);
docs_honesty-pinned sentence + (con Int) snippet preserved verbatim.
- T8: regression gate — workspace green; null-ctx fallback
byte-behaviour preserved.
- T9: bench-trio — compile_check/cross_lang clean; check.py
tracked-P2 noise only, causally exonerated (zero codegen/runtime
diff vs HEAD), NO baseline ratified.
Boss-verified directly (not on report trust): docs_honesty_pin 5/0,
design_schema_drift 8/0, embed_export_hash_stable 1/0,
embed_export_gate 6/0, embed_e2e 1/0, embed_staticlib_alloc_guard
2/0, print_no_leak_pin 1/0 + e2e rc_stats 4/0; the tsan coherent-stop
proven by hand — per-ctx clean (exit 0, 0 warnings), shared-ctx a
real ThreadSanitizer data race at rc.c:157/:208 (exit 66).
Known debt (journal + roadmap): DESIGN.md '## Embedding ABI (M1)'
header still says (M1) while the body mirrors M2 — out of Task 7's
scope-limited regions by design; post-audit doc-honesty tidy.
Milestone-close audit next.
Tasks 1-4 land fully review-green and are the cohesive shippable
subset (the per-thread embedding ABI, fully wired + proven):
- T1: FIXED-FIRST alloc-guard baseline pin (RED captured -> GREEN at T4).
- T2: runtime/rc.c gains ailang_ctx_t {alloc_count,free_count} +
ailang_ctx_new/_free + __thread __ail_tls_ctx; the two increment
sites become 'if (_ctx) _ctx->... else g_rc_...'. g_rc_* statics +
ailang_rc_stats_atexit + the constructor RETAINED VERBATIM as the
null-ctx (single-threaded executable) fallback. rc_accounting_tsan.c
+ driver: per-ctx 8x200000 tsan-clean (de-globalisation positive proof).
- T3: codegen Target::StaticLib forwarder gains a leading 'ptr %ctx'
+ one '@__ail_tls_ctx = external thread_local global ptr' decl +
TLS save/store/restore around the BYTE-UNCHANGED internal call;
_adapter/_clos + internal arg vector untouched (M1 decision held);
doc-comment provisionality narrowed to the value/record layout.
- T4: build_staticlib rejects --alloc != rc (RC-only swarm artefact).
Boss adjudication of the orchestrator's Task-5 BLOCKED (spec-defect,
correctly surfaced not hacked): swarm.c's -DSHARED_CTX negative
control is structurally impossible — examples/embed_backtest_step.ail
is a non-allocating scalar kernel that never writes a ctx field, and
__ail_tls_ctx is __thread, so a shared-ctx scalar swarm has no
shared-memory write to race on (the spec's OWN item-1 honesty point).
The de-globalisation teeth belong to the item-1 rc_accounting harness
(shared ctx genuinely races on ctx->alloc_count; orchestrator
independently confirmed tsan exit 66). Spec amended in lockstep
(Goal coherent-stop, must-fail-axis item 2, Testing item 3,
Acceptance) so item 3's negative control is item-1's by the
de-globalisation-proof / capability-demo split; plan Task 5
restructured (swarm.c per-ctx capability demo only; negative-control
standing test relocated into the item-1 harness driver) + Task 3's
plan-transcription error (@ail_backtest_step ->
@ail_embed_backtest_step_step) corrected. This is a Boss
consistency-repair of an internally-contradictory clause whose intent
is already correctly realised in the same spec — not a redesign; no
brainstorm bounce. Task-5 working files discarded (corrected plan
reproduces them; a structurally-RED test must not enter main).
INDEX.md + final journal land at iter completion (re-dispatch [5,9]).
Restores the Decision-5 IR-readability affordance for main-free
kernels: ail emit-ir gains --emit=staticlib, symmetric with
ail build, reusing the M1-audited Target::StaticLib path via a new
one-line lower_workspace_staticlib convenience. Zero-export guard
byte-identical to build_staticlib's. DESIGN.md widened (not
narrowed) + a pre-existing M1 synopsis omission corrected. 3 new
E2E tests; no fixture minted, no doc pin (E2E is the coverage).
Plan 03493c9.
Resolves the M1-fieldtest [friction] + [spec_gap]#1 shared root:
form_a.md stated an unconditional 'every (fn ...) param needs an
(own/borrow) mode' rule in 4 places, contradicting shipped checker
behaviour (scalars take and require bare (con Int); a mode on a
scalar trips body-pointing use-after-consume/consume-while-borrowed).
All 4 sites rewritten symmetric to the existing return-type carve-out;
DESIGN.md gains the bare-scalar export-param rule + a corpus-grounded
Form-A snippet; RED-first anti-regrowth pin via FORM_A_SPEC + norm().
Zero language/checker/codegen change. Plan 4c266a6.
Surface-touch fieldtest of M1's (export "<sym>") + --emit=staticlib
+ scalar/effect-free gate. DESIGN.md + public examples only (no
compiler source). The M1 thesis is empirically substantiated:
- [working] Int EMA + Float leaky-integrator both first-try clean
end-to-end (author -> --emit=staticlib -> C link -> typed scalar
return; s==67 / s==1.875, exit 0). Clause-1 confirmed.
- [working] Both clause-3 rejections precise + self-correcting
(export-non-scalar-signature on IntList param;
export-has-effects on !IO). Clause-3 confirmed.
- [working] (export) orthogonality + zero-export staticlib guard
match spec.
- [friction] form_a.md 'wrap every param in own/borrow' misdirects
the headline scalar export -> body-pointing use-after-consume /
consume-while-borrowed; only bare (con Int) checks. Pre-existing
form_a.md defect M1 made acute (not introduced).
- [spec_gap] form_a.md item 1 unconditional; scalar params take no
mode (shares root with the friction).
- [spec_gap] no public emit-ir --emit=staticlib though Decision 5
makes kernel-IR readability load-bearing.
Boss verified independently via the public CLI (positive E2E builds
both archives; both rejections fire the documented codes). 0 bugs:
M1 capability + gate sound. Findings route to the roadmap as
follow-up (not M1-blocking, no debug).
Tasks 1-3 of docs/plans/embedding-abi-m1.1.md, a verified known-good
subset (cargo build --workspace exit 0; full workspace green; the
roundtrip_cli all-examples gate green):
- Task 1: CLI build-path MissingEntryMain baseline pin
(examples/embed_noentry_baseline.ail + embed_missing_main_baseline.rs;
triple-assertion, layered on the green codegen-unit pin lib.rs:3314).
- Task 2: additive FnDef.export: Option<String> (serde default +
skip_serializing_if, modelled on FnDef.doc); compile-driven thread
of export: None across ~85 FnDef/AstFnDef literals / 18 files incl.
all in-source #[cfg(test)] mod tests + drift/hash/spec-drift
literals + the codegen lib.rs:3320 in-source pin; hash-stability
golden pin; DESIGN.md fn-JSON "export" line. DONE_WITH_CONCERNS
(plan symbol content_hash_fn -> def_hash per hash_pin.rs, plan-
pseudo-vs-reality class, plan corrected).
- Task 3: Form-A (export "<sym>") modifier — parse_export +
parse_fn thread + write_fn_def emission + form_a.md grammar;
byte-identical round-trip.
Task 4 BLOCKED (Boss plan-defect: fixtures declared module != file
stem; AILang's loader hard-enforces module==stem at workspace.rs:438,
so ail check panicked on load before the gate). Boss resolution
(Option 2, overriding the orchestrator's Option 1, rationale in the
journal + plan Design-decision 6): keep descriptive embed_* filenames,
rename fixture MODULES to their stems — the C symbol backtest_step
stays via (export ...), demonstrating spec Decision 2's mangling-
decoupling rather than violating it; archive becomes
libembed_backtest_step.a, internal symbol @ail_embed_backtest_step_step,
host.c unchanged. Plan fixed (Design-decision 6 + Task-4 module==stem
+ Task-5/6/7 dependent strings + the no-*. Float-head + content_hash_fn
concerns folded). Headline fixture corrected; blocked-Task-4 WIP
discarded (recreated by the [4,7] re-dispatch from the fixed plan).
PARTIAL commit (not the iter's final commit): the implement Repair
path mandates committing the known-good subset so the [4,7]
re-dispatch starts from a clean tree that includes Tasks 1-3 (Tasks
4-7 structurally depend on the schema field + surface). INDEX line
added when the full iter lands post-re-dispatch.
User-authored proposal added during the embedding-abi-m1.1 implement
run (path-disjoint from the iter; Boss roadmap maintenance, not
bundled into the iter commit). Role-split of the 2905-line DESIGN.md
into a typed contract ledger + per-model whitepapers + decision-
records rehomed to docs/journals/, on the consumer/lifetime axis.
Brainstorm complete for Embedding ABI M1 (first of the M1–M5 arc).
User-approved 2026-05-18. Grounding-check PASS on 3rd dispatch
(dispatch 1 false-BLOCK on an in-source unit test it missed;
dispatch 2 PASS + located the pin; dispatch 3 PASS after honest
prose correction).
7 design decisions pinned, all derived from the M1–M5 arc except
the one genuine fork the user decided:
- additive FnDef.export: Option<String> (FnDef.doc precedent)
- (export "<sym>") author-chosen, mangling-decoupled C symbol
- lib<entry>.a (program only) + separate libailang_rt.a
- no runtime lifecycle API — documented no-call contract
- --emit=staticlib suppresses @main + MissingEntryMain; needs >=1 (export)
- export-signature gate: scalar-only + effect-free (clause-3 discriminator)
- Task 1 fixed: MissingEntryMain build-path baseline pin (RED-first)
Roadmap M1 flipped [ ] -> [~]. Next: planner in a fresh session
(user-chosen session shape).
The data-server-driven goal: a compiled AILang kernel callable
in-process from a concurrent Rust host so a backtest swarm runs
each backtest as an independent single-threaded computation.
Decomposed into five chronologically-bounded milestones (M1
linkable artifact+scalar entrypoint, M2 per-thread ctx+concurrency
safety, M3 frozen value layout+ADT crossing, M4 List-ADT sequence
crossing, M5 ail-embed adapter+swarm fieldtest), each its own full
brainstorm->plan->implement->audit cycle. Shared motivation + three
audited invariants (clean core / single-threaded per computation /
native-AOT-only) carried on the M1 entry. The flat array/slice
primitive is recorded P2 as a measured performance follow-up
gated on M5's fieldtest, not a capability gap — keeping the ABI
arc's scope clean. Decision made in the 2026-05-18 data-server
analysis chat; no spec yet (brainstorm produces M1's spec at start).
Architect clean (one [medium] advisory Sweep-5 self-match wart filed
as P3, not a fix iteration). Bench causally exonerated: HEAD vs
5bb7211/de66eb7 — all 5 firing bench binaries cmp-byte-identical, ail
sha256 identical at all 3 commits; the check.py firings are the two
tracked-P2 environmental classes (*.bump_s anchor staleness; *.max_us
-n5 tail-jitter, 4th on a zero-runtime-change milestone), NOT this
milestone — NO baseline ratify (Iron Law forbids ratifying noise).
compile_check 0/24, cross_lang 0/25 EXIT 0. No fieldtest (zero
authoring-surface change). docs-honesty-lint CLOSED: roadmap [~]->[x],
WhatsNew appended, INDEX line added.
Single-iteration milestone (mirror effect-doc-honesty). Two-pronged
tense+modality discriminator codified present-tense as a DESIGN.md
meta-subsection (Approach A, user-ratified); procedural enumeration =
discriminator applied to the full architect_sweeps.sh + new Sweep 5
output (not a frozen orchestrator list — the sweep already surfaces
post-mortem sites the architect has been waving through as legitimate
quotes); anti-regrowth = enumerated docs_honesty_pin.rs (hard
cargo-test gate, incl. protected-exception present-anchors against
over-correction) + Sweep 5 + ailang-architect.md "DESIGN.md honesty
drift" bullet (advisory standing forward check). No language/checker/
codegen/runtime change. grounding-check PASS (10/10 ratified).
Post-audit downstream-LLM-author field test of the post-removal
Form-A surface. 4 fresh real-world programs written from DESIGN.md +
form_a.md + public examples only (no compiler source): running
sum-of-squares accumulator (loop/recur → 385), grade cascade
(let-threaded if → 4/2/0), Horner polynomial straight-line build-up
(let-shadow chain → 73), multi-state bracket scanner threading
(rest,depth,ok) by tail recursion → true/false. 0 bugs, 0 friction,
0 spec_gap, 4 working — every task clean and correct on the first
try with only let/if/loop/recur; all parse|render|parse
byte-identical; zero mut/var/assign tokens produced (a doc-faithful
author did not reach for the removed construct).
The closest-to-friction case (multi-state machine restates the full
state tuple per tail-call) is correctly classified working, not a
spec_gap: that explicit-dataflow cost is the local-reasoning pillar
working as designed; mut's implicit cross-iteration persistence is
exactly the failure class the removal eliminates. mut-keyword
rejection is fail-closed with a diagnostic that enumerates the
surviving forms (no tombstone, no-nostalgia, as spec'd).
Removal thesis empirically CONFIRMED — mut was redundant with
let/if/loop/recur in 100% of the fresh tasks. The
remove-mut-var-assign milestone is fully ratified and CLOSED:
spec+plan+iter, audit clean (architect [high] form_a.md fixed in
.tidy, bench causally exonerated), fieldtest clean. Roadmap P0
flipped [~]→[x]; WhatsNew user-facing entry appended.
Milestone-close audit for remove-mut-var-assign (48e7774..07f0802).
Architect drift: one [high] — crates/ailang-core/specs/form_a.md
still documented the full mut/var/assign construct (EBNF 286-288,
prose 308-328, dangling `like mut` at 332). The spec named it for
deletion; the iter missed it. Root cause = the recon-undercount
class (4th recurrence) plus a Boss error: a plan-time grep checked
only docs/form_a.md (dir-scoped), got "No such file", and that
unverified non-existence claim propagated into the recon brief —
the real file is crates/ailang-core/specs/form_a.md. A stale doc
never compile-fails so the implement compile-sweep could not catch
it. Fixed inline as a Boss .tidy (CLAUDE.md context-already-loaded
carve-out; full review discipline kept — tree-wide verified zero
construct residue except the roadmap entry naming the milestone
itself; grammar now reads coherently reuse-as -> loop -> recur).
Bench causally exonerated: HEAD vs 48e7774 bench binaries
cmp-byte-identical (the -2520-line removal alters zero bytes of the
firing fixtures' machine code). check.py exit 1's 3 firings are the
tracked P2 bump_s environmental staleness + single-sample -n5
max_us jitter (explicit_at_rc.max_us +108% -> -2.30% ok on
identical-code rerun). PRISTINE met on the merits; NO baseline
ratify (Iron Law + spec foreclosure).
Roadmap: P0 refreshed (audit clean, fieldtest remains); the
ailang-plan-recon P2 escalated in place to a structural
recon-contract defect (2 named gaps + concrete fix); a new distinct
P2 filed for the check.py *.max_us -n5 3-for-3 false-positive.
mut/var/assign removed from AILang entirely and atomically. Deleted:
Term::Mut/Term::Assign/struct MutVar; the three Form-A keywords +
parse_mut/parse_assign + grammar EBNF; the 4 mut CheckError variants;
the mut_scope_stack synth threading (param dropped from synth + every
internal/external/test caller); the two lower_term arms; and every
exhaustive no-_ Term::Mut/Term::Assign match arm across 17 source
files — cut in lockstep with DESIGN.md, fixtures, the drift trio,
carve-out and roadmap so the schema is honest at every commit. No
catch-all wildcard introduced (verified). loop/recur + let/if are
the surviving forms.
The shared codegen alloca machinery survives (loop reuses it):
mut_var_allocas renamed binder_allocas (representation-only, loop
codegen byte-identical) and the shared Term::Lam escape guard
simplified to !loop_stack.is_empty() with the loop half
(LoopBinderCapturedByLambda) byte-equivalent. Feature-acceptance
applied inverted: the removed feature fails clause 2 (redundant)
and clause 3 (IS the iterated-mutable-state bug class).
Behaviour preservation is executable: mut_counter/mut_sum_floats
still print 55 after the faithful let/if rewrite. The removal is
made executable by the new mut_removed_pin.rs (4 must-fail pins).
Independent verification: cargo test --workspace 605/0, zero
residual mut symbols in any crate source, loop/recur non-regression
all green (55 / 500000500000 / infinite-compiles / the
lambda_capturing_loop_binder pin), roundtrip_cli PASS.
One DONE_WITH_CONCERNS: a 4th recurrence of the recon-undercount
class (in-source mod tests + a drift-pin fn + 5 orphaned mut
.ail.json carve-outs + a non-enumerated E0599); all resolved within
implementer remit, no behaviour change. Milestone-close audit then
fieldtest remain.
spec docs/specs/2026-05-18-remove-mut-var-assign.md (grounding PASS)
plan docs/plans/remove-mut-var-assign.1.md
One iteration (not the 4-then-3 split the brainstorm carrier
assumed): the recon's cross-references prove a Rust exhaustive-match
enum-variant removal is atomic and the workspace test suite is
necessarily red between the Rust cut and the fixture/doc/drift
cleanup, so the full cargo-test green gate can only sit at the final
task. 6 tasks: T1 RED must-fail pins, T2 the atomic Rust cut
(~40 match arms + mut_scope_stack signature + all ~37 callers +
escape-guard loop-half-preserving simplification + codegen rename
mut_var_allocas→binder_allocas; gate cargo build 0-errors + T1
GREEN), T3 dangling test-pin deletions, T4 DESIGN.md hard-delete,
T5 faithful let/if fixture rewrites + 3 rejection-probe deletions,
T6 carve_out 18→12 + roadmap + final full-suite + loop/recur
non-regression gate. Plan-recon resolved 3 open Qs (form_a.md does
not exist — grammar lives in parse.rs doc-comment; round_trip.rs +
roundtrip_cli.rs auto-sweep examples/; escape.rs:533 walker is
loop-reachable).
Removal milestone: Term::Mut + Term::Assign + the MutVar surface decl,
one indivisible feature. loop/recur (shipped, closed) + let/if cover
every shipped use; the construct fails the feature-acceptance criterion
applied inverted (clause 2 redundant, clause 3 IS the iterated-mutable-
state bug class). Approach A: no deprecation window, JSON tags fail
closed as generic unknown-variant, codegen alloca machinery kept (loop
reuses it) renamed binder_allocas, shared Term::Lam escape guard keeps
its loop half. Decoupled from Stateful-islands. Grounding-check PASS —
all 7 load-bearing assumptions ratified by currently-green tests.
Roadmap: mut-removal as in-flight P0; the DESIGN.md/docs honesty-lint
queued as a distinct P1 milestone (depends on the removal closing),
its own brainstorm — no aspirational Wunschdenken / no post-mortem in
DESIGN.md; reader sees the language as it IS now.
Architect drift review clean (zero drift/debt, diff-verified: single
Term::Loop write_term hunk, no non-render site / Term::Recur / doc /
lockstep touched). Bench: compile_check 0/24 + cross_lang 0/25 exit
0; check.py's latency-tail firings proven single-sample jitter by an
identical confirmatory re-run (p99.9 +37%→+10% ok, max +114%→+22%
ok); the sole persistent firing is the pre-tracked P2 *.bump_s
environmental staleness, causally impossible to attribute to a
projection-only ailang-prose change. All items carry-on — no fix, no
baseline ratify. Fieldtest not applicable (no authoring-surface
change). Roadmap P0 flipped to closed; WhatsNew user-facing entry
appended (user present → notification suppressed per policy).
The Term::Loop arm of write_term rendered binders as bare
`name = init;` statements inside the `loop { … }` body block, which
reads as C/Rust re-init-every-iteration semantics — a projection
lying about a body the prose surface exists to let the reader trust.
Rewrite to a parenthesised init-list on the keyword line,
`loop(acc = 0, i = 1) { … }`, positionally isomorphic to the
unchanged Term::Recur arm. Projection-only: loop/recur AST, Form-A,
JSON-AST, typecheck, codegen, and the Form-A↔JSON round-trip
invariant untouched. RED-first via two new committed
examples/*.prose.txt byte-equality snapshots + two snapshot_loop_*
tests; full ailang-prose suite 10/0; both ail-prose CLI byte-matches
green. Single-iteration milestone, brainstorm→plan→implement.
Spec: docs/specs/2026-05-18-prose-loop-binders.md
Plan: docs/plans/prose-loop-binders.1.md
Single-iteration, RED-first plan for the prose-loop-binders spec.
Task 1 pins the desired Option-A bytes in two new committed
.prose.txt fixtures + two snapshot tests (RED against the old
renderer). Task 2 rewrites the Term::Loop arm of write_term and
gates GREEN via the snapshot suite + two ail-prose CLI byte-matches
+ an exact untouched-set check. Recon-verified line anchors; planner
Step-5 self-review clean on all 8 items.
Form-B prose renders loop binders as bare `name = init;` statements
inside the `loop { … }` body, which reads as C/Rust re-init-every-
iteration semantics — a lie about a body the projection exists to
let the reader trust. Redesign: parenthesised init-list on the
keyword, `loop(acc = 0, i = 1) { … }`, positionally isomorphic to
`recur(...)`. Projection-only; loop/recur AST, Form A, JSON-AST,
typecheck, codegen, and the Form-A↔JSON round-trip invariant are
untouched. Step-7.5 grounding-check PASS. Roadmap entry opened P0
in-flight.
Post-audit downstream-LLM-author field test of the shipped
loop/recur surface (DESIGN.md + public examples only). 3 real
iterative programs (Newton isqrt, Collatz, Euclidean gcd) + 5
plausible-mistake negatives + 2 no-termination probes, all run
through the public ail CLI. 0 bugs. 4 working findings on the
milestone's own axes: rejection diagnostics point-exact AND
self-fixing; recur tail-position threads through match/let/outer-if
(spec only showed if); loop composes as a value sub-expression +
byte-stable round-trip; no-termination boundary exact. This
empirically substantiates the "LLM author can now write iterative
programs" claim.
Two orthogonal non-blocking findings, neither in loop/recur scope,
both routed to P2 todos (refused the scope creep into a loop/recur
tidy): niladic (app f) spec_gap independently re-confirms the
existing mut-local-F3 roadmap item (the design-fork decision
deliberately NOT auto-ratified under /boss — parked,
priority-strengthened); module-level (doc) diagnostic-hint friction
(one-line tidy). Boss-verified independently (gcd->27;
recur-outside-loop fires exact).
The standalone loop/recur milestone is fully ratified and CLOSED:
3 iterations + tidy shipped, audit clean (drift resolved, bench
pristine carry-on), fieldtest clean on every axis. Roadmap P0
marked closed.
GREEN side of the RED audit-trail commit 39380d3. The loop/recur
milestone-close audit found a [high] correctness defect — a lambda
capturing a loop binder passed `ail check` then panicked
unreachable!() at codegen (crates/ailang-codegen/src/lambda.rs:102)
on type-correct input. Fixed symmetrically to mut.4-tidy: new
CheckError::LoopBinderCapturedByLambda (code
loop-binder-captured-by-lambda, bracket-free F2 Display), the
Term::Lam escape guard gained a parallel loop_stack pass. Minimal
mechanism: loop_stack element Vec<Type> -> Vec<(String,Type)> so
the already-threaded per-loop frame carries binder names; recur
still reads .1 by position (Boss-call-2 positional invariant
preserved verbatim — the name is a second field for the escape
guard only, a consumer iter-2 did not anticipate). Codegen
byte-unchanged (typecheck-only fix). carve_out 17->18 + ct1-F2 +
DESIGN.md note lockstep.
Folded in (Boss-side, audit resolution): the two [medium]
doc-honesty edits the audit surfaced (mut_var_allocas rustdoc now
states its mut-var+loop-binder dual use; Term::Loop doc-comment
now describes the real shipped state, not iter-1's stale
"per-binder phi" forward-look) + a [low] P2 roadmap todo
(plan-recon undercount countermeasure, pairs with planner Step-5
items 7+8). Bench gate: all 3 scripts exit 0, 25 metrics 0
regressed — pristine, carry-on (no baseline/ratify).
cargo test --workspace 619 -> 622 / 0 red (Boss-reran
independently, hash_pin 11/0). The loop/recur milestone is
audit-clean; fieldtest is the only remaining step before close.
Milestone-close audit (architect, edd2558) surfaced a [high]
correctness defect: a (lam ...) body that references a loop binder
passes `ail check` (exit 0) then panics unreachable!() at codegen
(crates/ailang-codegen/src/lambda.rs:102) on type-correct input —
a DESIGN.md "Robustness against hallucinations" violation.
Root cause (debugger-confirmed): the Term::Lam escape guard at
crates/ailang-check/src/lib.rs:3608 inspects only mut_scope_stack;
loop binders live in `locals` (Term::Loop arm ~:3911, iter-2
design), so the mut.4-tidy MutVarCapturedByLambda-class rejection
has no loop-binder counterpart — the capture is accepted at check
and crashes codegen.
This RED commit pins the FIXED contract: the program must fail
`ail check` with code `loop-binder-captured-by-lambda` (symmetric
to mut-var-captured-by-lambda), NOT panic codegen. RED today
(left: [] vs right: ["loop-binder-captured-by-lambda"]); 6 prior
loop_recur_typecheck_pin tests still green. GREEN follows.
Third and terminal iteration of the standalone loop/recur
milestone (plan eae73bf). Replaces the iter-1 lower_term
CodegenError::Internal stub for Term::Loop/Term::Recur with real
LLVM-IR lowering: loop binders as entry-block allocas (mut.3
pending_entry_allocas, reusing mut_var_allocas so the existing
Term::Var load path is byte-unchanged), a fresh loop-header block,
recur stores + back-edge br, a loop_frames stack saved/restored at
the lambda boundary. clang -O2 mem2reg promotes the allocas to
phi. Four Boss design calls implemented verbatim and journalled
(alloca-not-hand-phi; mut_var_allocas reuse; emergent loop-exit
via the if/match join; single block_terminated field + parallel
SET site). Diff confirmed surgical: codegen/lib.rs 3 hunks (field
+ init + stub->2-arms), lambda.rs 2 hunks (save+restore); zero
edits to any existing block_terminated SET/READ site, tail-app
lowering, or verify_tail_positions (Boss call 4, the spec-pinned
invariant). Three new .ail fixtures: sum_to->55,
deep-n 1e6->500000500000 (clause-2 correctness made executable),
infinite-loop build-only. Codegen-only: no schema/typecheck
change; hash pins + drift trio stay green untouched.
One DONE_WITH_CONCERNS (T3): the plan's `cargo test ... tail`
filter resolved to no tests; ran via real names + the full 619/0
which subsumes it (feedback_plan_pseudo_vs_reality class, no
behaviour change). Boss systemic fix folded in: planner SKILL.md
Step-5 gains item 8 (verification-command filter strings must
resolve) — the second planner-meta-gap this milestone surfaced.
cargo test --workspace 616 -> 619 / 0 red (Boss-reran
independently); the 3 loop/recur e2e explicitly green. All three
components shipped: the loop/recur milestone is structurally
complete. Milestone-close audit + fieldtest is the next step.
Terminal iteration of the standalone loop/recur milestone.
Replaces the iter-1 lower_term CodegenError::Internal stub with
real LLVM-IR lowering: loop binders as entry-block allocas (the
mut.3 pending_entry_allocas mechanism, reusing mut_var_allocas so
the existing Term::Var load path needs zero new code), a fresh
loop-header block reached by an unconditional br, recur stores +
back-edge br, a new loop_frames codegen stack saved/restored at
the lambda boundary (mut.3 triple precedent). Four Boss design
calls baked into the plan header, all on architectural-consistency
/ spec-pinned-invariant grounds (not effort): (1) alloca + clang
-O2 mem2reg, NOT hand-emitted phi (no linear-emit phi precedent;
spec's "phi" is the secondary impl-shape; mem2reg gives identical
optimized output); (2) reuse mut_var_allocas + the existing
Term::Var load path (Var-lowering byte-unchanged; representation-
sharing only); (3) loop-exit value is the emergent product of the
existing if/match join once recur sets block_terminated, no
separate result phi; (4) reuse the single block_terminated field,
recur sets it at its own emit site (parallel setter), zero edits
to existing SET/READ sites (tail-app byte-unchanged). Three new
.ail fixtures: sum_to->55 run, deep-n 1e6->500000500000 (clause-2
correctness made executable), infinite-loop build-only. iter-3 is
codegen-only: no schema/typecheck change, hash pins + drift trio
stay green untouched. Self-review item 7 (the systemic fix folded
into iter-2) applied: T1's field-add is not a fn-signature change,
no deferred-caller-past-gate.
Second of three iterations of the standalone loop/recur milestone
(plan 5ac57fe). Replaces the iter-1 synth CheckError::Internal stub
for Term::Loop/Term::Recur with real binder typing + positional
recur arity/type checking via a new loop_stack: &mut Vec<Vec<Type>>
frame threaded as mut.2's mut_scope_stack, a new private
verify_loop_body tail-position pass (sibling of the byte-frozen
verify_tail_positions — 0 deletions there), and the four Recur*
diagnostics firing point-exactly on four negative fixtures. The
iter-1 loop_sum_to.ail fixture now also typechecks clean; an
infinite loop typechecks (no termination claim). NO codegen (the
iter-1 lower_term stub stays — iter 3); NO Diverge/guardedness
(spec boundary).
Three Boss design calls implemented verbatim and journalled:
RecurTypeMismatch is an Assign-style structural pre-check (not a
unify-propagate); loop_stack is positional Vec<Type> (binder names
via ordinary locals); diagnostic-code precedence is by pass
ordering (no explicit logic).
Two DONE_WITH_CONCERNS, both journalled: a plan-ordering defect
(Task 2's compile gate forced the check_fn threading the plan
deferred to Task 4 — resolved byte-identically, only resequenced)
and the recurring mut.2-class recon-undercount of cross-module
synth callers (resolved via the plan's compile-sweep oracle).
Boss systemic fix folded in: planner SKILL.md Step-5 gains item 7
(compile-gate vs. deferred-caller ordering) so the plan-ordering
defect class is scrubbed at plan time. cargo test --workspace
608 -> 616 / 0 red (Boss-reran independently).
Second of three iterations of the standalone loop/recur milestone.
Replaces the iter-1 synth CheckError::Internal stub with real
binder typing + positional recur arity/type checking via a
loop_stack frame, a new private verify_loop_body tail-position
pass (sibling of the spec-frozen verify_tail_positions), and the
four Recur* diagnostics. Three Boss design calls baked into the
plan header: (1) RecurTypeMismatch is an Assign-style structural
subst-compare not a unify-propagate (so the dedicated code fires
point-exactly per spec acceptance); (2) loop_stack element type is
Vec<Type> (positional, not name-keyed IndexMap — recur rebinds by
position; binder names enter ordinary locals like Term::Let);
(3) diagnostic-code precedence is by pass ordering (synth before
verify_loop_body), already correct, no explicit logic needed.
Every code step literal — synth signature/threading, the two real
arms, verify_loop_body full body, four negative .ail.json
fixtures, carve-out 13->17 + stale-header fix, pin harness, ct1 F2
sibling — all verified against read source. Codegen stays the
iter-1 stub (iter 3). Self-review: reused iter-1's loop_sum_to.ail
as the positive-check fixture rather than create a near-duplicate.
First of three iterations of the standalone loop/recur milestone
(spec 97c1ed1). loop/recur become real parseable / printable /
round-trippable / hash-stable strictly-additive AST nodes across
every no-wildcard exhaustive Term match in all six crates, plus
parse/print, prose arms, DESIGN.md + form_a.md schema blocks,
drift/coverage/hash anchors, and the spec's worked sum_to program
as a green round-trip fixture. NO typecheck semantics and NO real
codegen this iter by design: synth and lower_term are stubbed
CheckError::Internal / CodegenError::Internal exactly as mut.1
(real typecheck = iter 2, real loop-header/phi/back-edge = iter 3).
Two binding Boss design calls recorded in the plan header and the
journal: (1) verify_tail_positions "byte-unchanged" = its tail-app
verification role is unchanged, not its source — it is an
exhaustive no-wildcard match so two descent-only arms are
mandatory; acceptance evidence is the tail-app non-regression test.
(2) codegen is in iter-1 scope as stubs/pass-throughs because the
workspace must compile for cargo test --workspace.
Three plan-pseudo-vs-reality substitutions (serde Type::Con
literal, bare Int -> (con Int), unqualified LoopBinder) and one
recon-undercount integration-test site, all intent-preserving and
journalled. Boss forward-fix folded in: the committed specs
themselves wrote bare Int in the loop-binder snippets (parses as
Type::Var) — corrected the three headline snippets to (con Int) for
doc-honesty (no design/schema change).
cargo test --workspace 600 -> 608 / 0 red (Boss-reran
independently); iter13a + new loop_recur hash pins green.
First iteration of the loop-recur milestone: Term::Loop /
Term::Recur / LoopBinder as strictly-additive nodes across all six
crates' no-wildcard Term matches, parse/print/round-trip,
hash-stability + drift anchors. No typecheck semantics (synth
stubs CheckError::Internal — iter 2), no real codegen (lower_term
stubs CodegenError::Internal — iter 3); both stub points mirror
mut.1. Two Boss calls baked into the plan header: (1)
verify_tail_positions 'byte-unchanged' = tail-app role unchanged,
operationalised via the tail-app IR-snapshot non-regression test
(literal source freeze impossible — exhaustive no-wildcard match);
(2) codegen is in iter-1 scope as stubs/pass-throughs because the
workspace must compile for cargo test --workspace (the iter goal).
Recon DONE_WITH_CONCERNS — all three open questions resolved at
plan time (Q3: print.rs:352 is a Type match, untouched).
Brainstorm milestone spec for first-class additive Term::Loop /
Term::Recur (Approach A). De-bundled from the reverted
Iteration-discipline milestone: no structural-guardedness checker,
no Diverge effect — the half-enforced-dichotomy failure that sank
the original bundle is explicitly out of scope. recur has no
non-tail spelling (structural jump, by-construction stack safety).
verify_tail_positions left byte-unchanged; new private
verify_loop_body pass owns recur-tail-position checking. Feature
acceptance: all 3 clauses pass. Grounding-check Step 7.5: PASS,
9 ratified load-bearing assumptions. User-approved.
Distillation of the effects→input→state→closures→recursion design
session. Ratifies one meta-principle (disciplined surface, mechanism
stays in the compiler) as standing orchestrator guidance, separates
the three bundled DESIGN.md commitments (content-address-definitions
is cheap and kept; total value-purity + RC-no-collector are the
expensive ones), and parks four open forks with epistemic status for
later brainstorm pickup. No feature ratified here.
Brummel