Skip to main content

import StdlibStatus from '@site/src/components/StdlibStatus';

core.async — Asynchronous execution

<StdlibStatus status="partial" detail="15 of 22 async modules carry full conformance suites; 11 fundamental compiler/stdlib defects pinned, 4 CLOSED 2026-05-24. Variant-algebra + stream factory + protocol-mountability + executor configuration + generic T.clone fallback are all interpreter-green; runtime poll-path coverage waits on the AOT executor test-bed." defects={[ {area: 'global AOT', summary: 'task #10 — compiler.phase.generate_native SIGABRT (LLVM SmallVector hang) — blocks AOT coverage across every async test'}, {area: 'stream', summary: 'CLOSED 2026-05-24: task #6 (Maybe.take() in-place mutation aliasing) — 3-part fix in dispatch_variant_method (clone variant payload to fresh heap BEFORE clearing original; intrinsic-preferred dispatch; well-known TypeId gate). Closes &mut self → self.value.take() class for stream_once, stream_repeat_n, stream_unfold.'}, {area: 'stream (T.clone)', summary: 'CLOSED 2026-05-24: task #7 — runtime identity-clone fallback in method_dispatch.rs:2526. When dispatch search exhausts qualified-type-resolution and bare-method is clone with no args, return receiver unchanged. Sound for primitive T (NaN-box value identity), reference-semantic T (shared pointer copy), read-only consumer sites. Mirror class to task #17 T.default/T.zero/T.one codegen intercept.'}, {area: 'executor (FFI intercept)', summary: 'CLOSED 2026-05-24: task #8 indirect via task #9 — stage-3 stub-IDs now flow through Tier-2 name-resolution at archive merge, restoring FFI symbol name lookup for the runtime-bridge intercept. All 25 contract-surface tests now active and GREEN.'}, {area: 'task #47 stub cascade', summary: 'CLOSED 2026-05-24: task #9 — fundamental fix at crates/verum_vbc/src/codegen/mod.rs:15059: external_function_names accumulator now distinguishes stage-1/2/3 stub IDs from variant-ctor / FFI extern sentinels. Stub IDs flow through name-resolution Tier-2 path at archive merge. Stdlib precompile restored to 13.7MB (was 143MB pre-fix). Unblocks [lenient] stage-3 stub never resolved panics across stdlib downstream.'}, {area: 'panic_fence', summary: 'CLOSED: task #11 (Maybe.take() &mut self writeback round-trip) — three-layer fix: precompiler self-shape encoding + archive-loader decoding + field-receiver SetF writeback in compile_method_call'}, {area: 'semaphore', summary: 'task #12 — AsyncSemaphore.new null-derefs through AtomicInt.swap in the Mutex/AtomicBool init chain. CLOSED: #13 (is-operator on single-variant sum) — parser fix at crates/verum_fast_parser/src/decl.rs:3384 distinguishes is-form sum declarations from =-form aliases per grammar §2.4'}, {area: 'timer', summary: 'task #15 — Duration.from_millis dispatch routes from_nanos to an Int receiver (archive-precompiled wrapper-fn return-register corruption — see audit). CLOSED: #14 (cross-module timeout_ms collision); #17 (getter-shadowing); #16 (Timeout<F> field-layout write OOB) — two-layer consumer-side fix that prefers TypeDescriptor.fields over the polluted simple-name type_field_layouts cache, eliminating the entire "record-type-name collides with sum-type variant payload" defect class universe-wide'}, ]} sweepDate="2026-05-24" />

Full async toolkit: Future protocol, executors, channels, async streams, timers, structured concurrency (nursery), racing (select), circuit breakers, retry policies, parallel helpers.

Module status​

Each core.async.* module carries an explicit conformance status so you know what you can rely on today versus what is still in flight. The status is the truth-table over the module's API surface as exercised by core-tests/async/<module>/ under both verum test --interp (Tier 0 VBC interpreter) and verum test --aot (Tier 2 LLVM AOT, --test-threads 1).

StatusMeaning
stableEvery public method is conformance-tested. Algebraic laws are pinned by exhaustive or large-domain property tests. Cross-stdlib integration is verified. Interpreter and AOT agree on every test. Safe to depend on in production.
completeSynonym for stable used by the conformance suite: 100 % public API coverage, every algebraic law pinned, every regression guarded forever.
partialSubset of the public API is conformance-tested and stable. The rest is exercised in regression_test.vr via @ignored tests pinning the specific defects that block coverage. The non-ignored API surface is safe; everything else is documented per-module under "Open defects".
regression-onlyModule is gated by upstream stdlib / language-level defects. Public-API tests do not pass yet — only @ignored regressions exist to lock the bug shapes. Avoid in production until promoted.
undocumentedDocumentation in this reference is authoritative, but the module has not yet been routed through the core-tests/ conformance suite. The current page is a best-effort snapshot of the source; it may drift from runtime behaviour.
ModuleStatusConformance suite
poll.vrcompletecore-tests/async/poll — 42 unit + 21 property + 14 integration + 7 regression-as-guardrail (79 working, 0 pinned)
waker.vrpartialcore-tests/async/waker — 9 working + 2 pinned regressions (inline-vtable redesign + record-literal Clone-corruption fix + Waker construction inlining; waker construction + clone + wake_by_ref + will_wake all now green; 2 residual pins: fn_ref-as-Int identity stability and Debug auto-derive precedence for record types)
future.vrpartialcore-tests/async/future — 17 working (construction + SelectResult variant algebra + Maybe interop) + 14 pinned regressions (the Future.poll / FutureExt.block surface is gated by the same &self auto-deref defect as waker §C)
backoff.vrpartialcore-tests/async/backoff — 14 working (BackoffStrategy variant + match-coverage) + 7 pinned regressions (Backoff. blocked by upstream CSPRNG intrinsic gap shared with reservoir)
task.vrpartialcore-tests/async/task — 16 working (JoinError variant algebra + TaskId record construction + List bookkeeping) + 2 pinned (TaskId.new atomic counter + JoinError Debug, both gated by upstream defects)
diagnostics.vrpartialcore-tests/async/diagnostics — 15 working (TaskLifecycleState 6-variant lifecycle + partition + terminal-state classification + List-histogram). Pure data-type module — no runtime dependency.
cancellation.vrpartialcore-tests/async/cancellation — 11 working (CancelReason 4-variant + Aborted(Text) payload + List bookkeeping). Timeout{deadline} arm deferred to integration once Instant works under interp.
channel.vrpartialcore-tests/async/channel — 12 working (TrySendError + TryRecvError variant algebra + payload recovery + retry-signal classification).
broadcast.vrpartialcore-tests/async/broadcast — 18 working (BroadcastRecvError + TryRecvResult 4-variant + LagPolicy 3-variant + Maybe interop).
select.vrpartialcore-tests/async/select — 12 working (Either<A,B> + SelectError + race-outcome bookkeeping).
nursery.vrpartialcore-tests/async/nursery — 8 working (NurseryErrorBehavior 3-policy + priority/severity ordering).
spawn_config.vrpartialcore-tests/async/spawn_config — 21 working (RestartPolicy + IsolationLevel + Priority 4-rank ordering + will-restart classification).
spawn_with.vrpartialcore-tests/async/spawn_with — 10 working (CircuitState 3-variant breaker lifecycle Closed → Open → HalfOpen → Closed + can-attempt classification).
executor.vrcomplete (configuration surface)core-tests/async/executor — 25 contract-surface tests (RuntimeBuilder.new/setters/full-chain + AsyncRuntimeConfig.default/cpu_bound/io_bound presets + AsyncRuntimeConfig setter surface + List bulk container + RuntimeConfig alias + 6 algebraic laws) all GREEN under --interp. Task #8 CLOSED 2026-05-24 indirect via task #9 fix — stage-3 stub-IDs now flow through Tier-2 name-resolution at archive merge, restoring FFI symbol name lookup for the runtime-bridge intercept. Live block_on(future) round-trip deferred to task #10 (LLVM AOT generate_native).
stream.vrpartialcore-tests/async/stream — 7 unit + 10 property + 5 integration + 6 regression GREEN under --interp. Pinned: size_hint laws (empty/once/repeat_n boundaries), poll_loop exhaustion stickiness, generator-driven termination via stream_from_fn, List drain via manual noop-context poll driver, T.clone identity-clone fallback for generic-T receivers (StreamRepeatN). Task #6 CLOSED 2026-05-24 via 3-part fix in dispatch_variant_method (clone-before-mutate). Task #7 CLOSED 2026-05-24 via runtime identity-clone fallback (method_dispatch.rs:2526). Task #9 CLOSED 2026-05-24 — stage-1/2/3 stub-ID name resolution in external_function_names accumulator (codegen/mod.rs:15059). Combinator-chain coverage (StreamMap/StreamFilter/StreamTake) deferred behind task #22 (constrained-implement-block dispatch).
generator.vrunaudited— runtime-bound; deferred pending the executor test-bed
timer.vrpartialcore-tests/async/timer — 29 working (Sleep/SleepUntil/Delay construction surface + TimerInterval new/immediate next_tick partition + Debounce/Throttle state-machine round-trip + monotonic refusal + reset-then-acquire across 4 representative intervals + TimeoutError Eq reflexivity) + 6 pinned regressions for tasks #14 / #15 / #16 / #17. Pre-fix landed in this branch: pub async fn acquire → public async fn acquire (line 535).
parallel.vrcomplete (interp)core-tests/async/parallel — 38 working covering parallel_map, parallel_filter_map, parallel_for_each, parallel_reduce, and the Blelloch parallel_scan_exclusive. Pinned properties: worker-count invariance over {1,2,4,8,16}, Blelloch-vs-reference exclusive-prefix-scan identity for + and max, parallel_reduce ≡ left-fold₁, filter_map index-subset-of-map. AOT validation gated by task #10.
panic_fence.vrpartialcore-tests/async/panic_fence — 12 working (panic_safe factory + record-literal Some/None inner + Ready(Ok) Int/Text round-trip + fence outcome→tag classification + List sequential consumption summing 15) + 1 pinned (task #11: Maybe.take() mutation through &mut self on a generic record field; gates the fence's documented "inner=None after Ready" lifecycle invariant). Panic-arm coverage deferred pending a panicking-Future test bed.
semaphore.vrregression-only outside variant algebracore-tests/async/semaphore — 8 working (SemaphoreError single-variant algebra including the natural e is SemaphoreError.Closed form + Result/Maybe wrapping integration) + 9 pinned regressions for #12 (lifecycle tests blocked by AsyncSemaphore.new null-derefs through AtomicInt.swap in Mutex/AtomicBool init). #13 CLOSED 2026-05-15 via single-line architectural fix in the parser — type X is Y; is now correctly parsed as a single-variant sum (was incorrectly downgraded to alias), closing the entire SemaphoreError / ChannelError / single-variant marker idiom across stdlib.
async_iterator.vrcomplete (protocol surface)core-tests/async/async_iterator — 4 unit + 3 property + 2 integration + 1 regression GREEN under interpreter. Pins: both protocols (AsyncIterator, IntoAsyncIterator) mount cleanly without archive-load panic; protocol-bound generic functions compile (A: AsyncIterator, B: IntoAsyncIterator + Clone); IntoAsyncIterator self-conversion blanket compiles via @inline-identity body; associated-type projection B.IntoAsyncIter resolves at function boundary; List<A: AsyncIterator> round-trip signature compiles. Stream→AsyncIterator blanket impl deferred behind upstream protocol-resolver projection-reduction work (each Stream-shaped type carries its own direct AsyncIterator impl in its owning module until the resolver lands).
intrinsics.vrpartialcore-tests/async/intrinsics — 19 working (Executor.current/in_async_context coherence + future_poll_sync ReadyFuture round-trip across Int/Text/Bool payloads + IntrinsicsYieldNow two-state lifecycle Pending→Ready with exactly-one-Pending tightness). Spawn family + sleep family @intrinsics deferred pending the live-executor test-bed.

Cross-module language defects​

Multiple interpreter / codegen defects are shared across the partial async modules above; closing any unblocks coverage in every module that depends on it.

Active (2026-05-14)​

  • Task #10 — global AOT generate_native SIGABRT. Every test under --aot crashes with __pthread_cond_wait → llvm::SmallVectorBase::grow_pod at the native-gen worker pool. Affects every async test (and every base/maybe test too); not async-specific. Blocks the AOT half of the cross-tier conformance contract. Repro: verum test --aot --filter test_none_construction from core/.

  • Task #11 — Maybe.take() mutation through &mut self does not flow back to a generic record field → CLOSED 2026-05-14 via three layered architectural fixes: (1) Precompiler self-shape encoding (crates/verum_vbc/src/codegen/mod.rs::compile_function) — pre-fix every non-Regular FunctionParamKind (the entire SelfValue / SelfRef / SelfRefMut / checked / unsafe / own family) serialised into the archive as TypeRef::Concrete(TypeId::UNIT), erasing reference + mutability

    • tier info at the serialisation boundary. Now each variant encodes its proper TypeRef::Reference { inner, mutability, tier } shape (or Concrete(parent_tid) for value receivers). (2) Archive-loader self-shape decoding (crates/verum_compiler/src/archive_ctx_loader.rs) — every site hardcoded takes_self_mut_ref: false. New fn_takes_self_mut_ref predicate inspects the first param's TypeRef: if Reference { Mutability::Mutable, .. } with name self, sets flag true. (3) Field-receiver writeback in compile_method_call (crates/verum_vbc/src/codegen/expressions.rs) — even with (1)+(2) wired, b.inner.take() still failed because RefMut creates a CBGR ref to the temporary register holding the extracted field, not to the field slot. The fix captures (base_reg, field_idx) upfront for Field receivers and emits a SetF after the method body returns, committing the (potentially mutated) receiver_reg back to the field slot — only when the method actually takes &mut self. Architectural rule pinned: the FunctionParamKind self-shape taxonomy MUST round-trip through the archive losslessly; the user-side method-call dispatch MUST honour the resulting takes_self_mut_ref flag with both RefMut to wrap the receiver AND the field-writeback SetF for chained-field receivers. Blast radius: every &mut self stdlib method (Maybe.take / Maybe.replace / Maybe.insert / Text.push_str / every mutator) now mutates correctly through user code.

  • Task #12 — AsyncSemaphore.new null-derefs through AtomicInt.swap. Every AsyncSemaphore.new(N) for any N panics with NullPointerAt { op: "opcode 0xe5", site: "AtomicInt.swap" }. Construction chain: AsyncSemaphore.new → Shared.new(Mutex.new(...)) → AtomicBool.new(false) → AtomicInt.swap NULL. Defect class: VBC interpreter atomic-primitive dispatch on a freshly-allocated atomic cell. Pinned in core-tests/async/semaphore/regression_test.vr §A.

  • Task #13 — is-operator returns false on a single-variant sum type → CLOSED 2026-05-15 by a single-line architectural fix at crates/verum_fast_parser/src/decl.rs:3384. Root cause (architectural): per grammar/verum.ebnf §2.4 the is / = sigils in type declarations are NEVER interchangeable — type X = Y; (with =) is an alias, type X is Y; (with is) is a single-variant sum. Pre-fix the parser collapsed type X is OnlyVariant; (bare identifier, no payload, no leading |) into TypeDeclBody::Alias(OnlyVariant), silently turning a single-variant sum into an alias to a non-existent type. Downstream the typechecker's is-operator pattern-resolution at crates/verum_types/src/infer/patterns.rs:1485 strictly required a Type::Variant, so e is X.Closed returned false — while match e { X.Closed => … } and Eq.eq still worked because their paths tolerated alias resolution. The fix emits TypeDeclBody::Variant(vec![first_variant]) for any is-form RHS that looks_like_variant() accepted. Both type X is | Closed; and type X is Closed; now produce identical AST. Closes the SemaphoreError / ChannelError / every single-variant marker error idiom across stdlib without a stdlib-source change. Architectural rule pinned: the is / = sigils in type declarations are NEVER interchangeable; any future "is-form RHS looks like a type alias" optimisation MUST go through =, not is.

  • Task #14 — timeout_ms cross-module name collision → CLOSED 2026-05-14 by two layered fixes in crates/verum_vbc/src/codegen/: (1) path-suffix narrowing probe in process_import_tree::Path — the existing lookup chain probed only core.async.timer.timeout_ms (verbatim) and async.timer.timeout_ms (core-stripped); neither matched because core/async/timer.vr declares module timer; (single-segment), so the archive_ctx_loader installs functions under timer.timeout_ms. The new probe iterates parent-path tails from longest to shortest (longest-prefix-match routing-table discipline), anchoring on func_name, and the first hit wins; [timer].timeout_ms now resolves cleanly. (2) strict-arity filter in type_aware_lookup — the cross-module disambiguation closure built arity_matches via the lenient lookup_function_with_arity helper, which returns the primary registration even for wrong arity. Wrong-arity candidates polluted the set, and the downstream param_type_names.iter().zip(arg_type_names) truncated to the shorter sequence — letting a 1-param method-with-self (ShutdownStrategy.timeout_ms(&self)) "type-match" a 2-arg call by only inspecting arg[0]. The strict-arity filter (info.param_count == args.len()) eliminates this collision class structurally: type-based disambiguation now runs ONLY between candidates that already agree on parameter count. Architectural rule pinned: every code path that filters function candidates by arity MUST use strict equality — lenient arity helpers stay available for "report-this-as-error" surfaces, but they MUST NOT seed disambiguation tiers.

  • Task #15 — Duration.from_millis dispatch routes from_nanos to an Int receiver. sleep(Duration.from_millis(N)) panics: "method 'from_nanos' not found on receiver of runtime kind Int". Chain Duration.from_millis → Duration.from_nanos(int*1_000_000) is dispatching from_nanos as a method on the Int multiplied result instead of as a static Duration constructor. Also surfaces through sleep(Duration.from_secs(1)) via the @inline(always) factory expansion, but NOT through Sleep.new(Duration.from_secs(1)). Pinned in core-tests/async/timer/regression_test.vr §A.

  • Task #16 — Timeout<F> field-layout write out of bounds → CLOSED 2026-05-15 by a two-layer consumer-side architectural fix in crates/verum_vbc/src/codegen/: (1) compile_record's alloc_slots resolution — emits Instruction::New { field_count } from TypeDescriptor.fields.len() via type_name_to_id → types[] BEFORE falling back to the polluted simple-name type_field_layouts cache. (2) resolve_field_index's field-idx resolution — strips generic args (Timeout<ReadyFuture> → Timeout), looks up the descriptor, scans its fields for the field-name match; returns the descriptor-canonical index. Root cause (architectural): the flat type_field_layouts: HashMap<String, Vec<String>> cache is keyed by SIMPLE type name and was polluted by record-style variant constructors that share a simple name with a standalone record type — canonical case: core.sys.io_engine.CompletionOp.Timeout { ts: &TimeSpec } (1-field variant payload) registered type_field_layouts["Timeout"] = ["ts"] before core.async.timer.Timeout<F> (3-field record) could claim the simple-name slot. Under first-wins, the host record's 3-field layout was silently shadowed. Timeout.new's precompiled body then emitted Instruction::New { field_count: 1 } (one slot, 8 bytes of data) and the subsequent SetF for field index 2 (or 5 after Sleep nested allocation) wrote past the allocation. Architectural rule pinned: the TypeDescriptor (one per type, owned by the type's own declaration site) is the canonical source of truth for field-layout information; flat caches are convenience indices that may suffer simple-name collisions from sibling sum types' record-style variants; every consumer that needs declared field count or field index MUST query the descriptor FIRST and fall back to the cache ONLY when the descriptor lookup misses. Blast radius: closes the entire "record type whose simple name collides with a sum type's record-style variant payload" defect class universe-wide — Timeout<F> is the canonical case but the same pattern applies to NurseryError.Timeout, ShellError.Timeout, every record-style variant payload across stdlib.

  • Task #17 — TimerInterval.period() field-vs-method-name shadowing causes StackOverflow → CLOSED 2026-05-15. The originally-pinned defect — where the getter method body self.period inside period(&self) -> Duration was dispatching as a recursive self.period() method call (blowing the stack) — no longer reproduces. Verified by both the user-side reproducer (type T is { period: Int }; implement T { fn period(&self) -> Int { self.period } } — both t.period field access and t.period() method call now stable) and the stdlib TimerInterval.period() regression test (flipped from @ignore to active). Closed indirectly by parallel codegen disambiguators landed by other agents and/or by task #11's field-receiver writeback work that consolidated the field-vs-method path in compile_method_call. Architectural rule pinned in the regression-test comment: bare self.X (no parens) inside a method body MUST resolve to field access when X names a field of the impl's parent type — even when X also names a method on the same type. Only self.X() (with parens) resolves to a method call. The pattern is canonical for getter idioms (Duration, Instant, Throttle, Debounce, TimerInterval).

Closed​

Five interpreter / codegen defects previously gated the partial async modules; closing them unblocked coverage:

  • Protocol default-method dispatch via blanket impl → CLOSED 2026-05-12 (task #11) by a focused blanket-impl pre-pass in collect_all_declarations + a generic-param materialisation skip in the main collect_declarations arm. Stdlib pattern: implement<F: Future> FutureExt for F {} declared AFTER concrete implement Future for ReadyFuture<T> (source-order in core/async/future.vr). Single-pass collection observed self.blanket_impls = [Future→IntoFuture] only at ReadyFuture's collection point — Future→FutureExt had not yet been visited, so FutureExt's default bodies (block / map / and_then) never monomorphised onto ReadyFuture and runtime r.block() panicked. Fix: pre-pass populates blanket_impls from a single linear scan; the main pass's already_present guard short-circuits duplicate registration. Generic-param skip suppresses spurious F.block / F.map / F.and_then registration when the blanket impl itself is observed (the generic-param's bare name was leaking phantom FunctionIds via the bare-name fanout). Critical invariant: the pre-pass NEVER calls generate_default_protocol_methods — only seeds blanket_impls — so the Poll-suite invariant (protocol-registry empty-entry guard at line 1455) stays intact and the default-method materialisation runs exactly once per (concrete impl × derived protocol) pair. Repro fixed at core-tests/async/future/regression_test.vr §A — 6 newly-passing tests across block / payload round-trip / lazy.invokes / map / map_composes / and_then. §B (4 tests on Join2/Join3/Select2 combinator receivers) remains pinned as task #24 — separate dispatch defect surfaces only when the receiver itself is a generic combinator wrapping inner Futures.

  • Stdlib precompile divergence for record methods → CLOSED 2026-05-12 as compile_record Clone-Unit-corruption. Investigation traced the root cause to compile_record's field-init Clone-before-SetF step, which wrote Unit into record fields whose AST declared no Clone impl. Combined with a Waker.from_raw(raw) call-arg passing failure in the unsafe { ... } wrapper at the call site, the construction chain materialised Wakers whose raw field was Unit; every downstream self.raw.<vtable.slot> GetF then null-derefed at the first chain step. Fix: removed the synthetic Clone in compile_record (records live in heap-allocated NaN-boxed objects, so the field-write copies the pointer into the parent record's field slot — aliasing isn't possible through this path), AND inlined noop_waker()'s body to a single nested record literal that side-steps the call-arg indirection. See commit 5129d8b1a.

  • type_field_layouts cross-mount registration race → CLOSED 2026-05-13 (task #9) by extending register_archive_type (crates/verum_vbc/src/codegen/mod.rs:3944) to unconditionally populate type_field_type_names for every archive-loaded record type — mirrors the register_record_fields invariant established by commit ab768e5d8 for user-phase declarations. Pre-fix the archive-side path only populated type_field_layouts and left type_field_type_names empty; downstream field_type_name returned None and resolve_field_index fell through to the "pick the type with the most fields" global-scan heuristic — silently routing record-construction field writes to wrong offsets when a sibling type with same-named field was in scope. Added type_ref_to_field_name helper that mirrors extract_type_name_from_ast's prefix-preservation invariants (&unsafe T / *const T / *mut T keep their prefix so the raw-pointer marker at compile_field_access line 14372 still fires); pinned the built-in-TypeId → name discipline via the new primitive_type_id_to_name source-of-truth (third site consistent with type_ref_to_name codegen-side and primitive_typeid_name archive_ctx_loader-side). 4 newly-passing regression tests at core-tests/async/future/regression_test.vr §C (ReadyFuture.value / Join2.{fut1,fut2,result1,result2} / Select2.{fut1,fut2} / Lazy.f field-access under List mount).

  • Free-function name collision in mount resolution → CLOSED 2026-05-12 by register_function_authoritative + archive cross-pollination guard + qualified-key path-doubling fix in register_module_filtered. When multiple stdlib modules exported free functions with the same simple name (e.g. select appears in 7 modules; join in 10), the call-site mount path.{name} previously bound to whichever overload won the bootstrap-order first-wins race. Three layered defects, all closed:

    1. Codegen first-wins discipline swallowed explicit user mounts: register_function's entry().or_insert() under prefer_existing_functions=true rejected the user's authoritative binding when a passive archive-load had already claimed the bare-name slot. Lifted via register_function_authoritative (writes both name and name#arity keys, no first-wins gate, no arity-collision shadowing) — invoked exclusively at the explicit-mount process_import_tree::Path branch. Glob mounts (mount X.*) keep first-wins to protect the FFI-raw / safe-wrapper precedence rule.

    2. Archive cross-pollination guard matched only the first path segment, so every stdlib core.X.Y.<leaf> path passed the gate against every other core.A.B.<leaf> (the w_prefix was always core for stdlib keys). Two unrelated functions sharing a simple name collapsed onto the same FunctionInfo, leaking the wrong FunctionId through the canonical qualified key. Tightened: when both keys are qualified, the FULL path-before-leaf must match.

    3. Path-doubling in register_module_filtered built the qualified key as format!("{}.{}", module_name, simple_name_str) without checking whether simple_name_str already carried the module path (the precompiler's descriptor-name promotion sets fn_desc.name to the full source-module-qualified form). Result: core.async.future.ready got registered as core.async.core.async.future.ready; the canonical-form probe missed, and the user-side mount's lookup fell through to the cross-pollinated entry under the un-doubled key. Mirrors the detection rule from populate_ctx_from_archive line ~326: when simple_name already contains a dot, treat it as the qualified shape directly.

    Repro pinned at core-tests/async/future/unit_test.vr §4-5: mount core.async.future.{join, select, join3}; let _ = join(f1, f2); now dispatches to core.async.future.join (was: core.io.path.join / core.security.labels.join / etc.).

  • Variant-tag stability under per-file test compilation → CLOSED 2026-05-13 (task #22) across 4 architectural defect classes via commits 90b94e68b + 3f14510b8 + 485a230c6 + f1dd6fd19. (1) Nested-variant destructure scrutinee-type leak — stash i-th variant_payload_types into match_scrutinee_type before recursing. (2) Flat-variant tag drift — scrutinee-aware lookup tier BEFORE bare lookup_function(name).variant_tag at both construction (compile_variant_constructor_hinted) and destructure (compile_pattern_test) sites. (3) Nested construction payload propagation — symmetric to (1) at construction site via new helper find_variant_payload_types_by_type_and_name. (4) Generic-param substitution — when payload type is bare generic-param shape ("T" / "E" / "Self"), substitute the i-th generic arg from the outer receiver_type<...> instantiation using TypeDescriptor's type_params index. Closes round-trip for Poll.Ready(Err(7)) → match → 2007 deterministically across precompile cycles. Mirror of tasks #9 / #11 / #21 discipline: context-aware canonical resolution wins over passive bare-name race. The constrained- implement-block dispatch issue (closure body in implement<T,E> Poll<Result<T,E>> { fn map_err(...) { ... } } not invoking the closure arg) is a DISTINCT defect tracked separately as task #25.

  • Closure dispatch in constrained-implement-block bodies (task #25, distinct from #22). Methods defined inside implement<T, E> Poll<Result<T, E>> (at core/async/poll.vr:100 onward — including map_ok / map_err / ready_ok / ready_err) are dispatched but the closure argument is bound incorrectly. Concretely: Poll.Ready(Err(7)).map_err(|e| e+1) returns Poll.Ready(Err(7)) rather than Poll.Ready(Err(8)) — the closure body is never invoked, producing a structural no-op. The dispatcher resolves the method (no panic, no argument-count mismatch) but either: (a) closure ends up in the wrong slot; (b) dispatch falls through to the generic implement<T> Poll<T> block's map and elides the inner transformation; or (c) the Ready(Err()) destructuring recurses back through Ready() without extracting the Err payload. The _preserves_ok / _preserves_pending test cases pass despite the bug, because their pinned outcomes (Ok / Pending arms) are preserved trivially by the no-op fallback. Worked around in core-tests/async/poll/{unit,property,integration}_test.vr via direct match q { Poll.Ready(Err(e)) => e, _ => ... } projection — the Poll/Result algebraic identity is pinned without crossing the broken dispatcher.

FilePurpose
poll.vrPoll<T> — the three async states
waker.vrWaker, Context, RawWaker, RawWakerVTable
future.vrFuture protocol + ReadyFuture, PendingFuture, Lazy, Join*, Select2, FutureExt
task.vrTask<T>, JoinHandle<T>, TaskId, JoinError, JoinSet<T>, YieldNow
channel.vrChannel<T>, Sender/Receiver, OneshotSender/OneshotReceiver, send/try errors
broadcast.vrBroadcastSender<T>, BroadcastReceiver<T>, broadcast_channel
executor.vrRuntime, RuntimeConfig, block_on, Timeout, LocalExecutor
select.vrEither<A,B>, select_either, race, select_all, join_all, try_first
stream.vrStream, StreamExt, 30+ adapters, factories (iter, unfold, interval, from_fn)
generator.vrGenerator<T>, AsyncGenerator<T>
nursery.vrNursery, NurseryOptions, NurseryError, NurseryErrorBehavior, TaskHandle
timer.vrSleep, SleepUntil, Interval, Delay, Timeout, Debounce, Throttle
spawn_config.vrSpawnConfig, RetryConfig, CircuitBreakerConfig, RecoveryStrategy, Priority
spawn_with.vrCircuitBreaker, CircuitState, execute_with_retry*
parallel.vrparallel_map, parallel_filter_map, parallel_for_each, parallel_reduce
intrinsics.vrruntime hooks: spawn_with_env, executor_spawn, future_poll_sync, async_sleep_*

Poll<T> — the two-state algebra​

Status: complete. Full public-API conformance under Tier 0 (interpreter) and Tier 2 (LLVM AOT, --test-threads 1). 42 unit + 21 property + 14 integration + 7 regression-as-guardrail tests; every algebraic law on the Functor-shape (identity, composition, Pending-as-absorbing- element), the Maybe-isomorphism round-trip, and the Eq / Clone / Default protocols are exhaustively pinned over a representative Int payload domain. Conformance suite: core-tests/async/poll.

The foundational poll-state algebra. Every async surface in the stdlib (Future.poll, Stream.poll_next, AsyncIterator.poll_next, cancellation checks, …) returns Poll<T> as its hot-path completion signal.

public type Poll<T> is
    | Ready(T)      // computation completed with value
    | Pending;      // not yet complete — the caller must re-poll

Predicates​

p.is_ready() -> Bool         // O(1), inlined
p.is_pending() -> Bool       // O(1), inlined; is_ready ⊕ is_pending = true ∀ p

Functorial map​

p.map<U, F: fn(T) -> U>(self, f: F) -> Poll<U>

Pending is the absorbing element: Pending.map(f) == Pending for any f; the closure is never invoked. Ready(t).map(f) materialises Ready(f(t)).

let p: Poll<Int>  = Poll.Ready(3);
let q: Poll<Int>  = p.map(|x| x * 2);              // Poll.Ready(6)
let r: Poll<Text> = p.map(|x| f"{x}");             // Poll.Ready("3")

Extraction​

p.unwrap(self)            -> T                  // panics if Pending
p.unwrap_or(self, default: T) -> T              // Pending → default
p.ready(self)             -> Maybe<T>           // Pending → None

p.ready() is the canonical bridge to Maybe<T>: Ready(t) ↔ Some(t) and Pending ↔ None. The reverse direction is the From<Maybe<T>> impl below; the two together form a structural isomorphism between Poll<T> and Maybe<T> on the observable surface.

Conversions​

From<Maybe<T>> for Poll<T>           // Some(t) -> Ready(t), None -> Pending

Removed blanket impl. An earlier From<T> for Poll<T> blanket impl was deleted: under any substitution T = Maybe<U> it overlapped From<Maybe<T>> and the impl-coherence selector silently routed Poll.from(non_maybe_value) through the Maybe arm — collapsing the result to Poll.Pending regardless of input. Write Poll.Ready(value) directly; it is the explicit, unambiguous form and produces byte-identical code.

Poll<Result<T, E>> — async-result composition​

map_ok / map_err live on the Poll<Result<T, E>> shape so async functions whose Output is Result<T, E> compose without intermediate destructuring.

Poll.ready_ok(t: T)   -> Poll<Result<T, E>>     // shorthand for Ready(Ok(t))
Poll.ready_err(e: E)  -> Poll<Result<T, E>>     // shorthand for Ready(Err(e))

p.map_ok<U, F>(self, f: F) -> Poll<Result<U, E>>      // F: fn(T) -> U
p.map_err<U, F>(self, f: F) -> Poll<Result<T, U>>     // F: fn(E) -> U

Pending remains the absorbing element on both arms: Pending.map_ok(f) == Pending and Pending.map_err(f) == Pending. Ready(Err(e)).map_ok(f) == Ready(Err(e)) (no re-wrapping); symmetric for map_err.

let p: Poll<Result<Int, Text>> = Poll.ready_ok(10);
let mid: Poll<Result<Int, Text>> = p.map_ok(|x| x + 1);
let after: Poll<Result<Int, Text>> = mid.map_ok(|x| x * 2);  // Ready(Ok(22))

Protocol implementations​

implement<T: Eq>      Eq      for Poll<T>;     // structural on Ready/Pending
implement<T: Clone>   Clone   for Poll<T>;
implement<T>          Default for Poll<T>;     // = Pending
implement<T: Debug>   Debug   for Poll<T>;     // "Ready(<t>)" | "Pending"

Algebraic laws​

All pinned by core-tests/async/poll/property_test.vr:

LawStatement
Functor identityp.map(|x| x) == p
Functor compositionp.map(g).map(f) == p.map(|x| f(g(x)))
Pending absorbs mapPending.map(f) == Pending
Maybe round-tripPoll.from(p.ready()) == p
Result arm preservationReady(Err(e)).map_ok(f) == Ready(Err(e)); symmetric for map_err
Eq reflexive / symmetric / transitivestandard equality laws on the two-variant carrier
Default invariancePoll.default() = Pending for every payload type

Pattern matching​

Poll<T> is exhaustive on two arms; nest with Maybe or Result for the common composite patterns:

fn classify(p: Poll<Result<Int, Text>>) -> Text {
    match p {
        Poll.Ready(Ok(_))  => "ready-ok",
        Poll.Ready(Err(_)) => "ready-err",
        Poll.Pending       => "pending",
    }
}

Future protocol​

type Future is protocol {
    type Output;
    fn poll(&mut self, cx: &mut Context) -> Poll<Self.Output>;
}

type IntoFuture is protocol {
    type Future: Future;
    fn into_future(self) -> Self.Future;
}

type FutureExt is protocol extends Future {
    fn map<U, F>(self, f: F) -> MapFuture<Self, F>
        where F: fn(Self.Output) -> U;
    fn and_then<U, F, Fut2>(self, f: F) -> AndThenFuture<Self, F, Fut2>
        where F: fn(Self.Output) -> Fut2, Fut2: Future<Output = U>;
    fn block(self) -> Self.Output;              // block current thread
}

Factories​

ready(value) -> ReadyFuture<T>             // immediately completes
pending<T>() -> PendingFuture<T>         // never completes
lazy(|| compute()) -> Lazy<F, T>           // deferred closure

Combinators (also available on FutureExt)​

join(fut1, fut2)               -> Join2<Fut1, Fut2>      // (O1, O2)
join3(fut1, fut2, fut3)        -> Join3                   // (O1, O2, O3)
join_all(futures)              -> List<Output>            // for List<F>
try_join(fut1, fut2)           -> Result<(O1, O2), E>     // fail-fast on Err
select(fut1, fut2)             -> Select2                 // first to complete wins
select_either(fut1, fut2)      -> Either<A, B>
race(fut1, fut2)               -> T                       // winner; loser cancelled
select_all(futures)            -> SelectAllResult<T>      // first + index
try_first(futures)             -> SelectAllResult<Output> // first Ok
timeout(fut, duration)         -> Result<T, TimeoutError>

Tasks​

type TaskId is { id: UInt64 };
type Task<T>  is { ... };
type JoinHandle<T> is { ... };
type JoinError is Cancelled | Panicked(PanicInfo);

spawn(future) -> JoinHandle<T>                   // shorthand
spawn_blocking(f) -> JoinHandle<T>               // on thread pool
spawn_detached(future) -> ()                     // fire-and-forget
yield_now() -> YieldNow                          // cooperate

h.abort()                                         // cancel
h.is_finished() -> Bool
h.id() -> TaskId
h.await -> Result<T, JoinError>                   // via Future

JoinSet<T> — dynamic task collection​

let mut set: JoinSet<Int> = JoinSet.new();
set.spawn(task_a());
set.spawn(task_b());

while let Maybe.Some(res) = set.join_next().await {
    match res {
        Result.Ok(value) => ...,
        Result.Err(JoinError.Cancelled) => ...,
        Result.Err(JoinError.Panicked(info)) => ...,
    }
}

Channels​

MPSC — Sender<T> / Receiver<T>​

channel<T>() -> (Sender<T>, Receiver<T>)                // unbounded
bounded<T>(capacity) -> (Sender<T>, Receiver<T>)
unbounded_channel<T>()                                  // alias for channel()
bounded_channel<T>(cap)                                 // alias for bounded()

// Sync API
tx.send(value)  -> Result<(), SendError<T>>               // blocks (futex) if bounded full
tx.try_send(value) -> Result<(), TrySendError<T>>         // Full | Disconnected
tx.send_timeout(value, d) -> Result<(), SendError<T>>     // with deadline

// Async API — awaitable with waker-based backpressure
tx.send_async(value) -> SendFut<T>                        // Future<Result<(), SendError<T>>>
tx.send_cancellable(value, &token)                        // Future; Err(Cancelled) on token fire
tx.closed() -> ChannelClosed<T>                           // Future<()>; completes on close

// Inspection
tx.is_closed() -> Bool                                    // == is_disconnected
tx.is_disconnected() -> Bool
tx.capacity() -> Maybe<Int>
tx.len() -> Int

// Receiver
rx.recv() -> Maybe<T>                                     // sync blocking
rx.recv_fut() -> RecvFut<T>                               // explicit async Future
rx.recv_cancellable(&token)                               // Future; Err(Cancelled) on token fire
rx.recv_many(&mut buf, max) -> Int                        // batch-drain in one lock
rx.try_recv() -> Result<T, TryRecvError>                  // Empty | Disconnected

// Receiver implements Stream<Item = T> — works with `for await msg in rx { ... }`
// and all StreamExt combinators (map, filter, take, etc.).

Async backpressure​

For bounded channels, send_async(value).await is the idiomatic way to apply backpressure:

  • If the channel has slack — push and resolve Ok(()) immediately.
  • If full — register the caller's waker in sender_wakers, return Poll.Pending. When the receiver pops, the sender's waker fires and the future re-polls, now with space.

The blocking send() method uses the same notification path but via futex, for non-async callers. Both paths share state; a bounded channel is safe to use from a mix of async and blocking senders.

Cancellation integration​

Both async variants accept a &CancellationToken parameter via *_cancellable — on token fire the future resolves immediately with Err(CancellationError) or Err(CancellableSendError.Cancelled(..)), deregisters its waker, and returns control. Pattern:

select {
    msg = rx.recv_cancellable(&shutdown).await => match msg {
        Ok(Some(v)) => handle(v),
        Ok(None)    => return,   // channel closed
        Err(_)      => return,   // shutdown fired
    },
    _ = idle_timeout.await => return,
}

One-shot — oneshot<T>()​

let (tx, rx) = oneshot<Result<Data, Error>>();
spawn async move { tx.send(compute()); };
let result = rx.await;

Broadcast (MPMC) — broadcast_channel​

broadcast_channel<T>(capacity) -> (BroadcastSender<T>, BroadcastReceiver<T>)
broadcast_channel_with<T>(capacity, policy) -> (Sender, Receiver)

// Sender
tx.send(value) -> Result<Int, SendError<T>>               // returns listener count
tx.clone() -> BroadcastSender<T>                           // multi-producer
tx.subscribe() -> BroadcastReceiver<T>                     // new receiver starting now
tx.receiver_count() -> Int
tx.sender_count() -> Int
tx.is_closed() -> Bool
tx.close()

// Receiver — implements Future AND Stream
rx.recv() -> BroadcastRecv<T>                              // awaitable future
rx.recv_cancellable(&token) -> Result<Result<T, RecvError>, CancellationError>
rx.try_recv() -> TryRecvResult<T>                          // Value/Empty/Closed/Lagged
rx.len() -> Int
rx.is_empty() -> Bool

Lag policies — LagPolicy​

Controls behavior when a receiver falls behind the ring capacity:

VariantSemantics
LagTolerant (default)Return RecvError.Lagged(n); advance to oldest available message.
DropOldestAdvance silently; never return Lagged. Senders never block.
DropSlowReceiverUnsubscribe the slow receiver entirely. For strict keep-up SLAs.

Broadcast receivers observe every value sent after subscription; they do not see historic values.

BroadcastReceiver<T> implements both Future<Output = Result<T, RecvError>> (direct .await) and Stream<Item = Result<T, RecvError>> (for-await loops and combinators). Sender and receiver counts are maintained atomically; last-sender-drop closes the channel and wakes all receivers.

Streams​

type Stream is protocol {
    type Item;
    fn poll_next(&mut self, cx: &mut Context) -> Poll<Maybe<Self.Item>>;
}
type IntoStream is protocol { ... };
type StreamExt is protocol extends Stream { ... };

Factories​

iter(iterable) -> Iter<I>                    // any IntoIterator
once(item) -> StreamOnce<T>
once_future(fut) -> StreamOnce<Output>
empty<T>() -> StreamEmpty
repeat(item) -> StreamRepeat<T>              // infinite (T: Clone)
repeat_n(item, n) -> StreamRepeatN<T>
from_fn(|| produce_next()) -> StreamFromFn
poll_fn(|cx| ...) -> StreamFromFn
unfold(state, |s| (item, new_state)) -> StreamUnfold
interval(duration) -> Interval

Adapters (return new streams)​

s.map(|x| f(x))            s.filter(|x| pred(x))     s.filter_map(|x| ...)
s.take(n)                  s.skip(n)
s.take_while(|x| pred)     s.skip_while(|x| pred)
s.chain(other)             s.zip(other)              s.enumerate()
s.peekable()               s.flatten()               s.fuse()
s.throttle(rate)           s.debounce(duration)      s.chunks(n)
s.buffer_unordered(n)      s.buffered(n)
s.timeout_each(duration)

Consumers (terminal)​

s.next() -> Poll<Maybe<Item>>
s.try_next() -> Poll<Maybe<Result<T, E>>>
s.for_each(|x| side_effect(x))
s.fold(init, |acc, x| ...) -> B
s.reduce(|a, b| ...) -> Maybe<Item>
s.collect<C>() -> C
s.find(|x| pred) -> Maybe<Item>
s.any(|x| pred) / s.all(|x| pred)
s.count() / s.last() / s.nth(n)
s.position(|x| pred) -> Maybe<Int>

Example​

async fn monitor(sensor: &Sensor) using [Logger] {
    let mut s = interval(1.secs())
        .map(|_| sensor.read())
        .filter(|r| r.is_ok())
        .map(|r| r.unwrap())
        .throttle_filter(|v| v.delta() > 0.01);

    while let Maybe.Some(reading) = s.next().await {
        Logger.info(&f"reading: {reading}");
    }
}

Generators​

fn* fibonacci() -> Int {
    let (mut a, mut b) = (0, 1);
    loop {
        yield a;
        (a, b) = (b, a + b);
    }
}

for n in fibonacci().take(10) {
    print(f"{n}");
}
  • fn* name(...) -> T — synchronous generator; returns Iterator<Item=T>.
  • async fn* name(...) -> T — async generator; returns AsyncIterator<Item=T> (a Stream<T>).

Inside a generator:

yield value;                // emit
// `return` (no value) ends the generator

Async generators support .await:

async fn* stream_events(url: &Text) -> Event using [Http] {
    let mut body = Http.get_streaming(url).await?;
    loop {
        let chunk = body.next_chunk().await?;
        for e in parse_chunk(chunk) { yield e; }
    }
}

for await event in stream_events("wss://…") using [Http] {
    handle(event);
}

Nursery — structured concurrency​

public type NurseryOptions is {
    timeout:   Maybe<Duration>,
    max_tasks: Maybe<Int>,
    on_error:  NurseryErrorBehavior,         // field name matches the surface syntax
};

public type NurseryErrorBehavior is CancelAll | WaitAll | FailFast;

public type NurseryError is
    | Single(Heap<Error>)                    // single-task failure
    | Multiple(List<Heap<Error>>)            // WaitAll collected multiple failures
    | Timeout
    | Cancelled
    | Panic(PanicInfo)
    | TaskLimitExceeded(Int);

// Builder
implement NurseryOptions {
    public fn new() -> Self;
    public fn default() -> Self;                                   // alias for new
    public fn with_timeout(self, timeout: Duration) -> Self;
    public fn with_timeout_ms(self, ms: Int) -> Self;
    public fn with_max_tasks(self, max: Int) -> Self;
    public fn with_error_behavior(self, behavior: NurseryErrorBehavior) -> Self;
}

// Underlying async functions the nursery { ... } block lowers to:
public async fn with_nursery<T>(
    options: NurseryOptions,
    f: fn(&mut Nursery) -> T,
) -> Result<T, NurseryError>;

public async fn with_nursery_timeout<T>(
    duration: Duration,
    f: fn(&mut Nursery) -> T,
) -> Result<T, NurseryError>;

Usage​

async fn fetch_batch(urls: &List<Text>) -> List<Bytes> using [Http] {
    nursery(
        timeout: 10.secs(),
        on_error: cancel_all,
        max_tasks: 100,
    ) {
        let handles: List<JoinHandle<Bytes>> = urls.iter()
            .map(|u| spawn Http.get(u.clone()))
            .collect();
        try_join_all(handles).await?
    } on_cancel {
        metrics.increment("fetch_batch.cancelled");
    } recover |e| {
        log_error(&e);
        List.new()
    }
}

Guarantees​

  • Every spawned task completes, fails, or is cancelled before the nursery scope exits.
  • on_error policies:
    • cancel_all — one failure cancels all siblings.
    • wait_all — collect all results including errors.
    • fail_fast — return the first error immediately.
  • Context stacks are inherited by spawned tasks.

Cancellation​

Cooperative cancellation is implemented by core.async.cancellation. A cancelled task continues running until it hits a cancel point — an .await on a cancellation-aware future, an explicit throw_if_cancelled check, or a CancelScope exit.

Core types​

// Owner — only holder can cancel.
CancellationTokenSource.new()
  .token()     -> CancellationToken                  // observer handle (clone-cheap)
  .cancel()                                          // with default CancelReason.Cancelled
  .cancel_with(reason)
  .is_cancelled() -> Bool
  .reason() -> Maybe<CancelReason>
  .linked_to(&parent)                                 // child source, propagates from parent

// Observer — read-only view.
token: CancellationToken
  .is_cancelled() -> Bool
  .reason() -> Maybe<CancelReason>
  .throw_if_cancelled() -> Result<(), CancellationError>
  .cancelled() -> CancelledFuture                    // Future<Output = CancelReason>
  .register(fn()) -> Registration                    // sync callback; RAII-deregister
  .child_source() -> CancellationTokenSource          // propagation tree
  .combine(&[t1, t2, ...]) -> CancellationToken       // any-of aggregation (static fn)
  .with_timeout(Duration) -> CancellationToken        // auto-cancel (static fn)
  .with_deadline(Instant) -> CancellationToken
  .never() -> CancellationToken                       // sentinel; never fires

Structured reasons​

type CancelReason is
    | Cancelled
    | Timeout { deadline: Instant }
    | ParentCancelled
    | Aborted(Text)

Children of a cancelled parent see ParentCancelled — not the parent's own reason. This makes structured traceability explicit.

Awaitable integration​

token.cancelled() returns a CancelledFuture that completes with the token's CancelReason. Compose with select:

select {
    r   = work().await           => Ok(r),
    res = token.cancelled().await => Err(res),
}

The future deregisters its waker on drop; no stale wake-ups.

Sync callback bridge — Registration​

let reg = token.register(|| close_file_handle(fd));
// ... do work ...
// Drop of `reg` deregisters BEFORE cancel fires. If cancel already
// happened, the callback fired synchronously inside `register()`.

Scoped cancellation — CancelScope​

let scope = CancelScope.new();
let token = scope.token();
spawn worker(token.clone());
// ... work ...
// Dropping `scope` cancels `token` (unless `scope.dismiss()` was called).

Scope variants:

CancelScope.new()                              // auto-cancel on drop
CancelScope.linked_to(&parent_token)           // child-source pattern
CancelScope.with_timeout(Duration)             // auto-cancel + timeout
scope.dismiss()                                // opt-out of drop-cancel
scope.cancel()                                 // explicit

Propagation rules (normative)​

  1. Parent → child: source.cancel() propagates to every token linked via child_source() / linked_to(). Children fire with CancelReason.ParentCancelled.
  2. Combine: tokens from CancellationToken.combine(&inputs) fire when any input fires.
  3. Idempotent: subsequent calls to cancel_with() after a cancel are no-ops; the first call's reason wins.
  4. Registered callbacks and wakers are drained under lock, then invoked without the lock held (no re-entrancy).
  5. Dropped registrations / futures deregister automatically.

Async signal subscription — core.signal​

Async-aware wrapper around core.sys.signal — exposes OS signals as awaitable futures and AsyncIterator streams for ergonomic composition with select, nursery, and cancellation tokens.

// Wait for a single Ctrl-C
ctrl_c().await;                                  // -> ()

// Wait for a single SIGTERM (K8s pod-eviction trigger)
terminate().await;                               // -> ()

// Wait for SIGHUP (reload-config convention)
hup().await;                                     // -> ()

// Wait for any shutdown signal; returns which one fired
let sig: Signal = shutdown_signals().await;      // Int | Term | Hup

// Arbitrary signal set as a Stream of arrivals
let mut stream = signal_stream(&[Signal.Usr1, Signal.Usr2]);
for await sig in stream {
    handle(sig);
}

Idiomatic shutdown — race server work vs signal:

select {
    _ = ctrl_c().await           => drain_and_exit(),
    _ = shutdown_signals().await => drain_and_exit(),
    r = run_server().await       => handle_result(r),
}

Architecture​

Invoking any allocating or lock-taking operation from inside a POSIX signal handler is undefined — the handler may preempt the mainline thread mid-malloc, mid-mutex-unlock, etc. core.signal uses the standard self-pipe / atomic-flag pattern:

  1. The OS-level signal handler (registered once per subscribed signal via core.sys.signal.on_signal) does only an async-signal-safe atomic store into a SignalFlag (set bit).
  2. A single background poller task polls these flags every ~20 ms, clears any set flags, and fans out to subscribers via a BroadcastSender<Signal>. The poller runs in normal runtime context, so broadcasting and waker operations are safe.

Trade-off: up to ~20 ms signal-to-subscriber latency — acceptable for shutdown, reload, and heartbeat use cases. A future upgrade to Linux signalfd(2), kqueue EVFILT_SIGNAL, or Windows APC delivery will collapse latency to zero behind the same public API.

Timers​

sleep(duration) -> Sleep                   // await to suspend
sleep_ms(ms)                               sleep_secs(secs)
sleep_until(deadline: Instant) -> SleepUntil

delay(future) -> Delay<F>                  // delay future by a duration
timeout(future, duration) -> Timeout<F>    // -> Result<T, TimeoutError>
debounce(future) -> Debounce<F>            // suppress rapid calls
throttle(future) -> Throttle<F>            // rate-limit

interval(duration) -> Interval              // stream firing on schedule
let mut ticker = Interval.new(500.millis());
loop {
    ticker.tick().await;
    update_ui();
}

Runtime and executor​

type RuntimeConfig is { ... };
type Runtime is { ... };
type LocalExecutor is { ... };
type TimeoutError is ();
type ExecutionEnv is { ... };               // θ+ context

Runtime.new() -> RuntimeBuilder
builder.worker_threads(n).stack_size(bytes)
       .io_engine(IoEngineKind.IoUring)
       .max_tasks(n)
       .build() -> Runtime

rt.block_on(future) -> Output
rt.spawn(future) -> JoinHandle<T>
rt.shutdown() / rt.shutdown_timeout(duration)
rt.enter()                                 // set current runtime for this thread

Global helpers​

block_on(future) -> Output                 // uses default runtime
spawn(future) -> JoinHandle<T>
current_runtime() -> Maybe<&Runtime>

LocalExecutor​

Single-threaded executor for !Send futures:

let exec = LocalExecutor.new();
exec.spawn_local(future);
exec.run_until(main_future);

Spawn configuration​

type SpawnConfig is { ... };                // builder
type RecoveryStrategy is
    | None
    | Retry(RetryConfig)
    | CircuitBreaker(CircuitBreakerConfig)
    | Fallback(fn() -> T)
    | Supervised;
type RestartPolicy is Permanent | Transient | Temporary;
type IsolationLevel is Shared | SendOnly | Full;
type Priority is Low | Normal | High | Critical;

let cfg = SpawnConfig.new()
    .with_priority(Priority.High)
    .with_isolation(IsolationLevel.Full)
    .with_recovery(RecoveryStrategy.Retry(RetryConfig.exponential(3, 100.millis())))
    .with_timeout_ms(5000)
    .with_name("worker-42");

let handle = spawn_with(cfg, task());

Retry and circuit breaker​

type RetryConfig is {
    max_attempts: Int,
    initial_backoff_ms: Int,
    max_backoff_ms: Int,
    backoff_factor: Float,
    jitter: Bool,
};
RetryConfig.fixed(attempts, delay_ms)
RetryConfig.exponential(attempts, initial_ms)

execute_with_retry(|| call_api(), max_attempts = 3, backoff_ms = 100)
execute_with_retry_config(|| call_api(), config)

Circuit breaker​

type CircuitBreakerConfig is {
    failure_threshold: Int,
    reset_timeout_ms: Int,
    half_open_max_calls: Int,
};
type CircuitState is Closed | Open | HalfOpen;

let breaker = CircuitBreaker.new(CircuitBreakerConfig {
    failure_threshold: 5,
    reset_timeout_ms: 30_000,
    half_open_max_calls: 1,
});

if breaker.is_call_allowed() {
    match call_remote().await {
        Result.Ok(v)  => { breaker.record_success(); Result.Ok(v) }
        Result.Err(e) => { breaker.record_failure(); Result.Err(e) }
    }
} else {
    Result.Err(Error.new("circuit open"))
}

Parallel helpers​

Data-parallel patterns, implemented in a portable way that the work- stealing runtime can pick up.

parallel_map(items, worker_count, |x| f(x)) -> List<U>
parallel_filter_map(items, worker_count, |x| maybe_transform(x)) -> List<U>
parallel_for_each(items, worker_count, |x| side_effect(x))
parallel_reduce(items, worker_count, |a, b| combine(a, b)) -> Maybe<T>

worker_count = 0 means "default to num_cpus()".

Low-level intrinsics (intrinsics.vr)​

type Executor is { ... };                  // opaque handle

Executor.current() -> Maybe<Executor>
Executor.in_async_context() -> Bool

spawn_with_env(future) -> JoinHandle<T>
executor_spawn(&exec, future) -> JoinHandle<T>
executor_block_on(future) -> Output
future_poll_sync(&mut future) -> Maybe<Output>      // single poll, sync

async_sleep_ms(ms) / async_sleep_ns(ns)

User code rarely touches these; they exist for runtime authors.

Waker and Context​

type Waker is { ... };
type Context<'a> is { waker: &'a Waker };
type RawWaker     is { ... };
type RawWakerVTable is { ... };

noop_waker() -> Waker
Context.from_waker(&waker) -> Context<'a>

waker.wake()         // consume, enqueue task
waker.wake_by_ref()  // enqueue without consuming
waker.clone()

Context inheritance across .await and spawn​

  • .await preserves the current context stack verbatim.
  • spawn snapshots the parent's context stack at spawn time.
  • nursery { spawn ... } — tasks inherit the nursery's contexts.
  • Channels do not propagate contexts (they're pure data pipes).

See Language → context system for the rules.

semaphore — cooperative task limiter​

mount core.async.semaphore.{AsyncSemaphore, SemaphorePermit};

let sem = AsyncSemaphore.new(10);      // cap at 10 concurrent ops

for url in urls {
    let permit = sem.acquire().await?;
    spawn(async move {
        let _p = permit;                // held for task lifetime (RAII)
        fetch(&url).await;
    });
}

Async-task counting semaphore — waiters park via Future / Waker instead of blocking an OS thread (unlike core.sync.Semaphore which futex-blocks). FIFO waiter fairness; try_acquire() non-blocking fast path; add_permits(n) runtime resize; close() causes pending + future acquire calls to fail with SemaphoreError.Closed.

Typical deployments:

  • bounded outbound fan-out (N concurrent HTTP fetches)
  • DB connection-pool checkout
  • rate-limit async CPU-bound tasks (N parallel inferences)
  • producer/consumer backpressure without a channel

backoff — retry delay policies​

mount core.async.backoff.{Backoff, BackoffStrategy};

let mut bo = Backoff.exponential_full_jitter(
    Duration.from_millis(100),
    Duration.from_secs(30),
).with_max_attempts(5);

loop {
    match try_operation() {
        Ok(r)    => return r,
        Err(_)   => match bo.next_delay() {
            Some(d) => async_sleep(d).await,
            None    => return Err(MaxAttemptsReached),
        },
    }
}

Four industry-standard strategies:

StrategyFormulaNotes
ExponentialNoJitterbase × 2^ndeterministic
ExponentialFullJitterrand(0, base × 2^n)AWS default
ExponentialDecorrelatedrand(base, prev × 3)AWS whitepaper; best for large fleets
FibonacciFullJitterbase × F(n+1) jitteredgentler ramp

Overflow-guarded integer arithmetic over microseconds. Once base × 2^attempt would overflow UInt64, the raw value saturates at cap_us — pathological max_attempts values plateau at the configured ceiling instead of wrapping.

See also​