Brummel
9339279181
Terminal iteration of the kernel-extension-mechanics milestone. Ships
the four language-level mechanisms named in the spec's § Goal:
Module.kernel + TypeDef.param-in schema, their Form-A surface,
flag-driven kernel-tier auto-injection, and generic param-in checker
enforcement with a new diagnostic.
Schema (Tasks 1+2). Module gains a `kernel: bool` field
(skip_serializing_if = is_false), TypeDef gains a
`param_in: BTreeMap<String, BTreeSet<String>>` field
(skip-if-empty, kebab-renamed to "param-in"). Both fields are
strictly additive — every pre-existing fixture's canonical-JSON hash
is bit-stable except `prelude.ail`, which intentionally gains
`(kernel)`. The struct-literal sweep covered ~104 Module sites and
~35 TypeDef sites across the workspace; the additive serde-default
covers JSON deserialise paths, only Rust struct literals broke.
Form-A surface (Tasks 3+4). `(kernel)` is a bare module-header
attribute; `(param-in (a Int Float) (b Str))` is one outer
TypeDef-body clause carrying one or more inner var-lists (OQ1
decision — mirrors `(ctors …)`, one parser arm, deterministic
BTreeMap iteration). Both round-trip Form-A → JSON → Form-A
bit-identical.
Workspace-load migration (Task 5). The hardcoded `&["prelude"]`
literal at loader.rs:108 became a `modules.values().filter(|m|
m.kernel)` derivation; `parse_prelude()` injection stays because
the prelude has no on-disk manifest in user workspaces. Prelude
now carries `(kernel)` in its source, so the new filter picks it
up automatically. Code-path migration only — observable behaviour
is identical (prelude_free_fns.rs stays green). prelude hash
re-pinned (af372f28c726f29f) with Honesty-Rule provenance comment.
WorkspaceLoadError::ReservedModuleName diagnostic prose
repurposed: any built-in kernel module name is reserved
(currently prelude + kernel_stub), not specifically prelude. CLI
mapping at main.rs updated in lockstep.
Stub crate (Task 6). New `crates/ailang-kernel-stub/` is a
zero-dependency leaf crate carrying only `pub const STUB_AIL:
&str` with the Form-A source of the kernel_stub module (one
parametric TypeDef with param-in, one ctor). The parse hop —
`parse_kernel_stub()` — lives in ailang-surface next to
parse_prelude, keeping the crate-dependency graph acyclic
(`ailang-surface → ailang-kernel-stub → ailang-core`, no
back-edge). The stub is injected unconditionally in all builds as
the ratifying fixture for the kernel-extension mechanism; future
base extensions may add more or retire the stub. Drift-pinned by
`kernel_stub_module_round_trips`.
Checker (Task 7). New `CheckError::ParamNotInRestrictedSet`
variant + code() + ctx() arms + enforcement in
`check_type_well_formed`'s Type::Con arm — generic, data-driven
from the TypeDef, mentions no specific extension type. Two
in-source tests pin both the rejection (`Str` outside `{Int,
Float}`) and the acceptance (`Int` inside) paths.
Workspace-load integration tests (Task 8). New
`workspace_kernel.rs` integration-test crate with three tests:
auto-import without explicit `(import …)` declaration, two
kernel-tier modules co-load, explicit-import-overrides-auto-
import precedence preserved. Loader is import-tree-only so the
auto-import tests use a bridge module that brings the kernel
module into the workspace via the import graph — docstring
captures the reachability nuance for future readers.
Doc-state transitions (Task 9). INDEX.md kernel-extensions row
annotation transitions from "design accepted 2026-05-28; impl in
progress" to "mechanisms milestone closed 2026-05-28; raw-buf and
series milestones pending". Whitepaper STATUS + auto-import +
param-in sections transitioned forward→present for shipped
mechanisms; forward-tense survives only in sections describing
the still-pending raw-buf/series milestones (per Honesty-Rule).
data-model contract gains anchor blocks for both new schema
fields.
Side-effect: every binary's IR snapshot now contains ~52 lines
for `drop_kernel_stub_StubT` because the stub is auto-injected
into every workspace load. Snapshots refreshed; e2e expects 4
modules per workspace (prelude + kernel_stub + entry + zero or
more user modules) instead of the previous 3.
Plan defects scrubbed in the implementation (folded back into
the planner template via the planner's self-review checklist
next time): Task 4 sample test src used fictional
`(ctors (MkT a))` list form (project grammar is per-`(ctor MkT
a)`); Task 6 original wiring would have created a cycle
ailang-surface → ailang-kernel-stub → ailang-surface (inverted —
stub crate is zero-dep, parse hop lives in surface); Task 7 in-
source tests referenced a fictional `check_type_in_module`
helper (used the existing Workspace + check_workspace
convention); Task 8 first integration test expected loader to
auto-load kernel modules from disk (loader is import-tree-only;
tests use a bridge module).
Concern-5 fix folded in pre-commit: workspace.rs ReservedModuleName
doc-prose initially said "in test/dev builds" for kernel_stub —
but stub is unconditionally injected in all builds. Doc copy
tightened to present-state per Honesty-Rule.
Stats: 0 spec-review-loops, 0 quality-review-loops, 2 sweep-script
retries on Task 2 (brace-depth bug on nested vec![Ctor{…}],
recovered via per-file checkout + rewritten anchor-on-existing-
field sweep), 1 e2e-snapshot refresh on Task 6.
81 lines
2.7 KiB
Rust
81 lines
2.7 KiB
Rust
//! Pin for `CheckError::DuplicateCtor` — same-named ctors in two
|
|
//! different ADTs within one module. Ratifies the load-bearing
|
|
//! consumer of the per-module `env.ctor_index` rebuild in
|
|
//! `check_in_workspace`. Env construction is code-SoT (the
|
|
//! `env-construction` ledger row in design/INDEX.md is source-link
|
|
//! only). The two overlays — typecheck-side rebuild + mono-side
|
|
//! narrowed — are deliberately asymmetric: the typecheck-side
|
|
//! `ctor_index` half exists *for this very diagnostic*, while the
|
|
//! mono-side overlay is narrowed to types-only because the runtime
|
|
//! ctor lookup is type-driven. If the per-module rebuild is ever
|
|
//! removed without an equivalent replacement path, this test goes
|
|
//! red.
|
|
|
|
use ailang_check::check_workspace;
|
|
use ailang_core::ast::{Ctor, Def, Module, TypeDef};
|
|
use ailang_core::workspace::{Registry, Workspace};
|
|
use ailang_core::SCHEMA;
|
|
use std::collections::BTreeMap;
|
|
|
|
#[test]
|
|
fn duplicate_ctor_across_two_adts_in_one_module() {
|
|
let m = Module {
|
|
schema: SCHEMA.into(),
|
|
name: "M".into(),
|
|
kernel: false,
|
|
imports: vec![],
|
|
defs: vec![
|
|
Def::Type(TypeDef {
|
|
name: "Cat".into(),
|
|
vars: vec![],
|
|
ctors: vec![Ctor {
|
|
name: "Twins".into(),
|
|
fields: vec![],
|
|
}],
|
|
doc: None,
|
|
drop_iterative: false,
|
|
param_in: BTreeMap::new(),
|
|
}),
|
|
Def::Type(TypeDef {
|
|
name: "Dog".into(),
|
|
vars: vec![],
|
|
ctors: vec![Ctor {
|
|
name: "Twins".into(),
|
|
fields: vec![],
|
|
}],
|
|
doc: None,
|
|
drop_iterative: false,
|
|
param_in: BTreeMap::new(),
|
|
}),
|
|
],
|
|
};
|
|
let mut modules = BTreeMap::new();
|
|
modules.insert(m.name.clone(), m.clone());
|
|
let ws = Workspace {
|
|
entry: m.name.clone(),
|
|
modules,
|
|
root_dir: std::path::PathBuf::from("."),
|
|
registry: Registry::default(),
|
|
};
|
|
|
|
let diags = check_workspace(&ws);
|
|
let dup = diags
|
|
.iter()
|
|
.find(|d| d.code == "duplicate-ctor")
|
|
.unwrap_or_else(|| {
|
|
panic!(
|
|
"expected at least one `duplicate-ctor` diagnostic, got: {diags:?}"
|
|
)
|
|
});
|
|
assert!(
|
|
dup.message.contains("Twins"),
|
|
"expected message to name the duplicate ctor `Twins`, got: {}",
|
|
dup.message
|
|
);
|
|
assert!(
|
|
dup.message.contains("Cat") && dup.message.contains("Dog"),
|
|
"expected message to name both ADTs `Cat` and `Dog`, got: {}",
|
|
dup.message
|
|
);
|
|
}
|