Static copy

Copied from docs/en/evidence/P00-G12-conformance-and-mvp-gate.md in https://github.com/poppy-team/aipo-lang (MIT). Pinned to revision ce47c4e070404fc41722ccd8e2b7b808b99062a8, blob e057ddc0397ff34514b6c19ff2f5a9bd1b6f54b1. The source repository remains canonical; this copy is refreshed through a sync pull request, not live.

Evidence — P00-G12 / Slice S11 (Conformance Hardening & MVP Gate)

Goal: P00-G12 — Conformance hardening: fixtures, snapshots, gauntlet and MVP gate Phase: P00 (Foundation) · Slice: S11 · Recorded: 2026-09-15 Environment: rustc 1.98.1 (48a229cea 2026-09-01), cargo 1.98.1 (797e8a9bc 2026-08-05), Linux

Proof attachments for the required gates. Commands are reproducible from the repository root. The corpus layout and snapshot matrix live in docs/conformance/README.md; this record is the score, not the specification.

Deliverable

The slice turns "the MVP works" into a claim a command can falsify. Everything the MVP subset claims is either executed with a committed output, or rejected with a committed diagnostic code:

Kind	Count	Location
Runnable programs + stdout snapshots	11 + 11	docs/conformance/programs/
Diagnostic fixtures + expected codes	10 + 10	docs/conformance/diagnostics/
Formatter golden pairs	8 + 8	docs/conformance/formatting/
Module cases	3	docs/conformance/modules/

Snapshots are observations and are regenerated only on request (AIPO_UPDATE_SNAPSHOTS=1); diagnostic codes, formatter expectations and module fixtures are hand-authored because they are specifications.

Gauntlet rubric

Gate	Weight	Command	Result
Executable MVP	30%	cargo test -p aipo-cli --test conformance	pass — 13/13, all program and module fixtures match committed stdout
Diagnostic discipline	20%	same suite, failure tests	pass — every fixture fails with its committed code, including the guard that a parse fixture reports an AIPO_PARSE_* code
Determinism	20%	cargo test -p aipo-formatter	pass — 11/11; formatter idempotent, canonical sources report no drift, fmt --check never rewrites
Robustness	15%	cargo test -p aipo-cli --test fuzz_smoke	pass — 3/3; random bytes, mutated programs and truncated programs never panic the pipeline
Repository gates	15%	cargo fmt --check, clippy -D warnings, test --workspace, doc	pass — all four green

Command output:

$ cargo test -p aipo-cli --test conformance
test result: ok. 13 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

$ cargo test -p aipo-cli --test fuzz_smoke
test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

$ cargo test -p aipo-formatter
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

Gate: fmt / check / clippy / test / doc

$ cargo fmt --all -- --check
FMT_EXIT=0

$ cargo check --workspace --all-targets
    Finished `dev` profile [unoptimized + debuginfo] target(s)

$ cargo clippy --workspace --all-targets -- -D warnings
(no output, exit 0)

$ cargo test --workspace
passed: 136  failed: 0

$ cargo doc --workspace --no-deps
   Generated /home/raillen/Documentos/Projetos/aipo-lang/target/doc/aipo_ast/index.html and 14 other files

MVP exit gate

docs/waves/wave-1-mvp.md lists five exit conditions. Status against each:

Exit condition	Status	Evidence
aipo run executes real small programs	met	programs/* snapshots, modules/* cases
aipo check reports diagnostics; aipo fmt idempotent	met	diagnostics/* fixtures; formatter golden + idempotency tests
pass/fail fixtures green; snapshots committed; integration tests green	met	conformance suite 13/13
No Rust panic escapes as user error (fuzz smoke)	met	fuzz_smoke 3/3
CI: fmt/clippy/test/doc green; prumo validate and prumo doctor green; docs delta resolved	met	gates above; prumo outputs below

$ prumo validate
Prumo validation passed.

$ prumo doctor
Prumo Doctor: all checks passed cleanly.

Verified gaps (gate is partial, and says why)

The corpus found four places where canon and the implementation still differ. Exercising them is the point of the corpus — the S11 gate does not hide them. Each is recorded in docs/adp/ADP-002-construction-hooks-and-runtime-contracts.md with a minimal demonstrating program, and each is carried as an acceptance criterion of P00-G13:

Gap	Today	Canon requires
G1 init not invoked by Type{...}	construction only fills named fields	construction runs init when declared
G2 invariant() not evaluated	instance published unchecked	invariants hold on publication
G3 signature contracts not checked at runtime	value crosses the boundary unchecked	runtime check at boundaries
G4 Bytes has no construction form	Bytes([1,2,3]) raises AIPO_RT_NOT_CALLABLE	construction and indexing are in scope

These are not regressions and not failures of this goal's acceptance criteria: S11's deliverable is the corpus and the honest gate, and it delivers both. They are the reason the gate is recorded as partial rather than closed.

Documentation delta

Artifact	Change
docs/conformance/README.md	new — corpus layout, fixture invariants, snapshot matrix, regeneration, gauntlet rubric
docs/adp/ADP-002-construction-hooks-and-runtime-contracts.md	new — the four verified gaps above, with demonstrating programs
docs/evidence/P00-G11-cli-and-formatter.md	records the S10 deliverables this slice scores
docs/evidence/P00-G10-backend-completion.md	records the backend gaps this slice exercises
.ai/goals/P00-G13.goal.json	successor goal for the four gaps
CHANGELOG.md, PROJECT_STATE.md	slice S11 recorded, next action set to S12