* async-spawn: build under no_std + alloc

* ci: guard the no_std contract for async + async-spawn with wasm32v1-none

* fmt: reorder imports in spawn.rs to satisfy rustfmt

…ables (#1601)

* rust: qualify world-owned type paths so future/stream payload vtables resolve

* add regression test to .wit file

* add ignores for wit test to cpp and go

* fix #1514 variant 3

* fix glitches in the patch

* fix temporary variable naming in deallocation

* Update tools used in CI

* Update Wasmtime to a fixed release
* Update wasm-tools to latest
* Update wasi-sdk to latest

* Go back to wasi-sdk-30, but parameterize it

This commit adjusts the behavior of the `async-spawn` feature by notably
ensuring that spawned tasks are polled even when the main task is
complete. This ensures that spawned work is looked at as opposed to
being stuck in limbo (never polled). This resolves an issue with
bytecodealliance/wasmtime#13196, for example, where work was spawned,
but the main set of work was complete, so the specific interleaving
there meant that the spawned work never actually ended up happening.

…at return strings, i.e. types returned via pointers. (#1606)

* add support for the ping-pong test.

Requires supporting futures that return strings.

* address feedback, remove redundant parameters and duplicate Callback function

* add example for regex

* feat(csharp): add map type support

Dictionary<K,V> provides a familiar, zero-dependency host representation
for WIT maps in C#, and unblocks runtime interop with components that
use them without requiring a custom collection type.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* fix(csharp): emit component_type.o for linker

The `map<K, V>` WIT syntax is a gated feature that the default wit-parser
used by wasm-component-ld's `--component-type` flag cannot parse, so the
text-based component type path was blocking any world that uses maps.
Switching to the same object-file approach already used by the C and C++
backends lets the component type flow through as a custom section without
round-tripping through the text parser.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* fix(csharp): embed component type via binary WIT package

Earlier attempt used `NativeFileReference` on the component-type object,
but NativeAOT's LLVM pipeline did not propagate the custom section into
the final module, leaving every C# component import unresolvable. Using
a binary-encoded WIT package (which `wasm-component-ld --component-type`
accepts alongside the text form) preserves the existing linker path and
also lets map<K,V> round-trip since decoding skips the default WIT
parser that rejects gated features.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* fix(csharp): force-link component-type object for NativeAOT

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* fix(csharp): defer map runtime tests until toolchain support lands

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* refactor(csharp): drop unused component-type object plumbing

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* chore(csharp): revert unrelated Cargo.lock drift

No Cargo.toml changed on this branch, so the lockfile should match main.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

---------

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

Keeping up-to-date.

* fix: add defer pinner.Unpin()

Signed-off-by: Andrew Steurer <94206073+asteurer@users.noreply.github.com>

* feat: add test

Signed-off-by: Andrew Steurer <94206073+asteurer@users.noreply.github.com>

---------

Signed-off-by: Andrew Steurer <94206073+asteurer@users.noreply.github.com>

* fix: remove inline wasm ABI pointer shifts

* docs: note MoonBit ABI layout change

Avoid lifting the raw core return value for async lifted exports without
a retptr. The core return is async progress information, while the
component result is delivered through task.return.

Additionaly params is now using &sig.results which makes AsyncTaskReturn
match the actual canonical ABI core signature: [], [i32], [ptr] etc.

* C++: fix compile errors when lowering `result<string>`, etc.

Previously, the code generator got confused about when to use `wit::string`
vs. `std::string_view` when lowering variants with string payloads,
tuple-with-string payloads, etc.

I'll admit my C++ skills are pretty out-of-date, so I'm not 100% confident I've
got everything lined up correctly here, but it fixes the scenarios I've run into
so far.  Feedback from experts in modern C++ would be very welcome.

* add `more-variants.wit-borrowed-duplicate` to `should_fail_verify` for Rust

* C++: sort types topologically before emitting definitions

Specifically, we sort the types while taking the parameter and return types of
resource functions into consideration.  This ensures that types are declared
(and, if necessary, defined) before they are used in the emitted source code.

Fixes #1615

* remove redundant resource type forward declarations

* use `wit_parser::TypeIdVisitor` to simplify `sort_types`

Classify fewer types as "primitive" to help get the names to line up.

Closes #1621

* feat(c): add map type support

The flat { entry_t *ptr; size_t len; } shape matches the canonical
ABI's lowered layout for map<K, V>, so reusing the list fast-path
keeps the FFI boundary copy-free and brings C to parity with the
Rust, Go, MoonBit, and C# backends that already ship map support.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

* test(c): add map runtime test variants

Covers the C-side lower/lift boundary behaviorally against the
existing Rust/Go/MoonBit suite, so future ABI changes that compile
cleanly cannot silently regress map round-trips.

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

---------

Signed-off-by: Yordis Prieto <yordis.prieto@gmail.com>

Rationale is listed in #1626.

Closes #1626

[automatically-tag-and-release-this-commit]

Co-authored-by: Auto Release Process <auto-release-process@users.noreply.github.com>