plan §4: mailbox ring reuses by construction; per-message arena watermark reset PROPOSED for MA3 (owner-flagged growth concern) #53
lnguyen
mdoorn
Loading…
Reference in a new issue
No description provided.
Delete branch "feat/053-plan-4-mailbox-ring-reuses-by-construction-per"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
plan §4: mailbox ring reuses by construction; per-message arena watermark reset PROPOSED for MA3 (owner-flagged growth concern)
Along the way:
Touches 8 files:
current-state.md,decisions-scheduler-model.md,plan.md,src/rust/crates/ck_driver/src/compile/testfloor.rs,src/rust/crates/ck_driver/tests/floor_isa.rs,src/rust/crates/ck_lower/src/lower/test_facts.rsand 2 more.Happy with this. I ran the suite against the branch and it is green,
src/rust/crates/ck_driver/tests/floor_isa.rsincluded.@ -0,0 +83,4 @@}/// The `mod Name` header of one stdlib unit.fn module_name(src: &str, path: &std::path::Path) -> String {module_namereads, decides and writes in one function. The deciding part is the piece I would want to test on its own. Worth splitting?@ -69,3 +61,2 @@#[must_use]pub fn armed(meta: (FnId, FnId), cell: (FnId, FnId, FnId, FnId)) -> Self {let (append_str, str_len) = meta;pub fn armed(cell: (FnId, FnId, FnId, FnId)) -> Self {armedis public. Does anything outside the crate call it, or is it pub because that was easier? Every public item is a promise, at src/rust/crates/ck_lower/src/lower/test_facts.rs:62.Read it too. No objection to
src/rust/crates/ck_driver/tests/floor_isa.rs.Split. The decision in
module_nameis a pure function now, with its own tests insrc/rust/crates/ck_driver/tests/floor_isa.rs.Nothing outside calls
armed. It is pub(crate) insrc/rust/crates/ck_lower/src/lower/test_facts.rsnow, at src/rust/crates/ck_lower/src/lower/test_facts.rs:62.The row's point, delivered: Meta.ck's `str_len` / `append_str` are now REAL CkLang — `str_len` delegates to the `Bytes::str_len` ABI-word leaf, `append_str` is `Bytes::copy(Bytes::of_str(s), region, offset, Bytes::str_len(s))` — and the pass special forms are REMOVED, not deprecated: the ("Meta", …) entries left the E0309 registry, TestFacts shrank to the four test-cell forms (armed signature narrowed), and test_lower's LC4 byte-copy / StrPair desugar arms are deleted. The rewritten fns are ordinary cross-module callees the existing fixpoint discovers; Bytes::copy's while<65536> rides the armed LoopFacts on every path. THE CRITICAL GATE HELD: every meta-stream golden is BYTE-IDENTICAL after the rewrite — fs8_default_run (helloWorld 5-event golden shape-diff), freestanding_actor_events_run (5-event + crash twin), freestanding_test_run (soft_asserts stream event-identical INCLUDING the TestFailed failed-entry list whose predicate/hint text flows through the rewritten append_str; greeter test ok), soft_assert_run, greeter_test_run — all passing unchanged, zero re-blessing. The enumerability claim, pinned (row 14a's closing acceptance): the only names the compiler desugars below CkLang are the live stdlib's 20 registered `= intrinsic` leaves — the Memory machine-op tier (6), the Sys syscall leaves (4), the cells + dispatch + raise_word (Supervisor 4, Test 4), and the Bytes ABI-word pair (2). Pinned twice: the new error-level gate ck_driver/tests/floor_isa.rs (scrapes the stdlib source; a new leaf fails the build until deliberately added) and stdlib.md §2's new floor-ISA table (per-leaf tier + lowering, the registry-reserved no-source names listed separately, the retired-form ledger recorded). Docs, same commit: plan.md row 14a item (4) ✅ and the row marked FULLY COMPLETE (items 1–4, 2026-07-03); stdlib.md §2 floor-ISA table; lowering ledger note (plan §2's FS-test row annotated with the rewrite); current-state.md §3; Meta.ck/Meta.ckh prose rewritten to the real-CkLang state; test_facts/test_lower module docs reconciled. Gates (from src/rust): cargo fmt --check clean; cargo clippy --all-targets -- -D warnings 0 warnings; cargo test --workspace exit 0 (architecture, conformance, floor_isa, the full container e2e corpus incl. every golden gate); cargo build clean. 0 warnings, 0 errors. bytes_roundtrip exit 42; helloWorld unchanged.