chore(deps): update wasmtime requirement from 41 to 45#2
chore(deps): update wasmtime requirement from 41 to 45#2dependabot[bot] wants to merge 1 commit into
Conversation
Updates the requirements on [wasmtime](https://github.com/bytecodealliance/wasmtime) to permit the latest version. - [Release notes](https://github.com/bytecodealliance/wasmtime/releases) - [Changelog](https://github.com/bytecodealliance/wasmtime/blob/main/RELEASES.md) - [Commits](bytecodealliance/wasmtime@v41.0.0...v45.0.0) --- updated-dependencies: - dependency-name: wasmtime dependency-version: 45.0.0 dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com>
|
Closing in favour of a combined wasmtime + wasmtime-wasi bump (see #1). Beyond the lockstep issue, v45 also moves |
|
OK, I won't notify you again about this release, but will get in touch when a new version is available. If you'd rather skip all updates until the next major or minor version, let me know by commenting If you change your mind, just re-open this PR and I'll resolve any conflicts on it. |
Combined replacement for the two split Dependabot PRs (#1, #2). The crates re-export types from each other and must move in lockstep; bumping only one produces 'mismatched types' between wasmtime::Error and wasmtime_internal_core::error::error::Error. Source changes required by v45: - wasmtime::Error no longer implements std::error::Error, so the anyhow::Context trait no longer applies to Result<_, wasmtime::Error>. Switch to wasmtime's own error::Context trait, which provides the same .context() method on those Results. - Config::async_support is deprecated and a no-op in v45 (async support is always enabled when the component-model feature is on). Drop the call.
Both snippets in the migration guide and docs/01 showed list<tuple<string, string>> shapes that don't match the shipped WIT. Adopters who copy the doc snippets get wit-bindgen type errors against the real interfaces. - chain::request-batch now uses the nominal record rpc-request and variant rpc-result (matching wit/nexum-runtime/chain.wit). - http now uses the nominal record header and adds the timeout-ms field (matching wit/nexum-runtime/http.wit), plus *.domain wildcard syntax in the allowlist example and the docstring on how non-2xx surfaces. Resolves PR #6 findings #2 and #3.
…lities) (#6) * docs: add 0.1 to 0.2 migration guide * wit: rename web3:runtime to nexum:runtime, unify error model, add identity/clock/random/http/query-module * chore: rename crate nxm-engine to nexum-engine; bump to 0.2.0 * build: update justfile and CI for nexum-runtime + nexum-engine rename * docs: rename to nexum:runtime, unify error model, mark non-server platforms as planned * runtime: update engine + example to 0.2 WIT - Engine main.rs targets world `shepherd` (formerly `shepherd-module`), generates against nexum:runtime@0.2.0 + shepherd:cow@0.2.0. - Replaces per-domain error records (JsonRpcError, MsgError, StoreError, ApiError, bare String) with unified HostError + HostErrorKind across every host impl. - Adds Identity host stub (was missing in 0.1 despite being doc'd). - Adds chain::request_batch stub that falls back to per-call dispatch. - Renames feed_get/feed_set -> read_feed/write_feed. - Drops separate cow + order interfaces, replaced by single cow_api with request and submit_order. - Block ts in test event is now ms (1_700_000_000_000) per types.wit documented unit. - Example module targets event-module world, matches Event::Tick instead of Event::Timer, returns HostError from init/on_event. * example: drop empty-name guard from init The guard was inconsistent: missing 'name' key silently defaulted to 'unknown' and returned Ok, but an explicit empty string returned Err — and both paths logged a success-shaped 'name=...' line BEFORE the guard ran. The example is a hello-world; a config-validation guard belongs in a real module, not a starter. Drop it. Resolves PR #6 finding #9. * ci: add wit deps sync check Both wit/nexum-runtime/ and wit/shepherd-cow/deps/nexum-runtime/ are committed to the tree; the latter is regenerated by 'just sync-wit'. Without a CI guard, a contributor can edit one without the other and the divergence only surfaces at runtime (interface-mismatch on module instantiation) — not at build time. The new job re-runs the same copy that 'just sync-wit' does, then fails if the deps tree differs from the freshly-copied canonical. Resolves PR #6 finding #8. * docs/01: align identity::sign to personal_sign semantics The shipped wit/nexum-runtime/identity.wit defines sign() as personal_sign — host MUST prepend the EIP-191 prefix ('\x19Ethereum Signed Message:\n<len>') before hashing. docs/01 described it as 'sign raw bytes' which is a transaction-signing footgun (a raw signer can be tricked into signing EIP-155 / EIP-712 payloads disguised as plain bytes) AND diverges from the WIT — two compliant hosts implementing different specs produce mutually unverifiable signatures. Align docs/01 to the WIT. Note that raw-bytes signing, gated by an explicit capability, is on the 0.3 roadmap. Resolves PR #6 finding #1. * docs: align chain::request-batch and http WIT snippets to shipped types Both snippets in the migration guide and docs/01 showed list<tuple<string, string>> shapes that don't match the shipped WIT. Adopters who copy the doc snippets get wit-bindgen type errors against the real interfaces. - chain::request-batch now uses the nominal record rpc-request and variant rpc-result (matching wit/nexum-runtime/chain.wit). - http now uses the nominal record header and adds the timeout-ms field (matching wit/nexum-runtime/http.wit), plus *.domain wildcard syntax in the allowlist example and the docstring on how non-2xx surfaces. Resolves PR #6 findings #2 and #3. * docs: defer typed config-value variant to 0.3 The shipped wit/nexum-runtime/types.wit ships 0.1's stringly-typed `type config = list<tuple<string, string>>`, but five doc locations plus the migration TL;DR promised a typed `config-value` variant for 0.2 (string/integer/boolean/list). Per architectural triage, defer the typed variant to 0.3 alongside the manifest parser work — the typing story lands as one coherent feature. Updates: docs/00, docs/01, docs/02 (two places), docs/08, migration guide TL;DR row + 'Typed config' subsection. Resolves PR #6 finding #4. * docs/migration: replace cargo-nexum vapor with real commands §9 verification checklist referenced 'cargo nexum check' and 'cargo nexum run --mock'; §11 promised 'cargo nexum migrate --from 0.1' as shipping with 0.2. None of these exist — there is no cargo-nexum crate in the workspace. Rewrite §9 to use real `cargo` + `just` invocations. Drop the §11 codemod promise; the sed cheat sheet in §8 already does the mechanical work. Add a clear 'no cargo-nexum toolchain in 0.2; coming in 0.3' note so adopters set the right expectation. Resolves PR #6 finding #7. * wit+engine: import clock/random/http in event-module; add host impls The new 0.2 capability interfaces (clock, random, http) shipped as standalone WIT files but were not imported by any world, so modules built against event-module / shepherd could not 'use' them. Their existence was advertised in the migration guide but was structurally unreachable. - event-module.wit: add 'import clock', 'import random', 'import http' (grouped as ambient host services after the six core primitives). shepherd world inherits via 'include event-module'. query-module remains intentionally sandboxed (no caps, no chain, no network). - Engine: add host impls for clock (SystemTime + monotonic_baseline: Instant), random (getrandom 0.4 fill), http (returns Unsupported for 0.2; real fetch is 0.3 — allowlist enforcement lands with #6). - Cargo.toml: add getrandom 0.4, plus serde 1 + toml 1 (for #6). Resolves PR #6 finding #5. * engine: minimal nexum.toml manifest parser with [capabilities] enforcement Per architectural triage, ship minimal manifest enforcement in 0.2 and defer optional-capability trap stubs to 0.3. The migration guide §3 promised four mechanisms; this commit lands the two security-critical ones (required-capability check + http allowlist enforcement) plus the deprecation warning, and explicitly defers per-import trap stubs. Manifest schema (parsed in crates/nexum-engine/src/manifest.rs): [module] name, version, component (sha256; parsed, not yet verified) [capabilities] required (validated against KNOWN_CAPABILITIES; engine rejects unknown names) optional (parsed + logged; trap-stub fallback is 0.3) [capabilities.http] allow (exact host or *.suffix wildcard) [config] TOML scalars flattened to strings (typed variant on 0.3 roadmap) CLI: nexum-engine <component.wasm> [<nexum.toml>]. If the second arg is omitted, the engine looks for nexum.toml next to the .wasm. If neither exists, it emits the deprecation warning and proceeds with empty config and empty allowlist (= effectively denies all HTTP). http::fetch now performs the per-module allowlist check (host extracted via a stdlib URL parser, exact or *.suffix match). Allowlist denial returns HostError { kind: Denied }; real network fetch is still 0.3. Includes unit tests for extract_host and host_allowed. Resolves PR #6 finding #6. * style: cargo fmt (nightly) CI rustfmt (nightly) collapses several multi-line signatures that the hand-formatted code had as multi-line. * rename WIT package nexum:runtime -> nexum:host (resolve engine/runtime naming overload) The 0.2 release shipped two similarly-named things — `nexum-engine` (the Rust crate hosting WASM components) and `nexum:runtime` (the WIT package). Both contain 'runtime' or a synonym, but they're at different layers: engine = implementation, runtime = contract. Worse, README.md described the engine crate as 'Host runtime — wasmtime-based component loader' — explicitly calling the *engine* 'runtime' while a sibling directory was literally named `nexum-runtime`. The word 'runtime' overwhelmingly means 'the thing that runs code' in programming usage; putting it on the contract side inverted the convention. Rename the WIT package to `nexum:host` — the host-imports surface a guest sees. Distinction becomes self-documenting: engine (nexum-engine) = a concrete implementation host (nexum:host) = the WIT contract every engine implements Touchpoints: - wit/nexum-runtime/ -> wit/nexum-host/ (12 files renamed via mv) - wit/shepherd-cow/deps/nexum-runtime/ -> wit/shepherd-cow/deps/nexum-host/ (regenerated by 'just sync-wit'; also covered by CI guard) - Every 'package nexum:runtime@0.2.0' -> 'package nexum:host@0.2.0' (12 files) - Every 'use nexum:runtime/...' -> 'use nexum:host/...' (shepherd-cow WIT) - Every 'nexum::runtime::...' -> 'nexum::host::...' (engine + example Rust) - justfile sync-wit paths - .github/workflows/ci.yml wit-deps-sync paths - All design docs (00..08), migration guide - README.md + docs/00 add an explicit 'Engine vs. host' callout so the distinction is directly apparent to a new reader. - A few stray 'host runtime' prose mentions in docs/05 and docs/07 changed to 'host engine' for consistency. Migration guide updated so a 0.1 -> 0.2 reader never encounters the mid-development 'nexum:runtime' name — the new package is `nexum:host` end-to-end. Build + run smoke clean; nightly rustfmt clean. * wit: drop deps/ vendoring; list both packages explicitly in bindgen The 0.2 release vendored a copy of wit/nexum-host/ under wit/shepherd-cow/deps/nexum-host/ because the engine's bindgen target (shepherd:cow/shepherd) imports from nexum:host via 'include', and wit-parser's default cross-package resolution looks at <pkg>/deps/. Maintained by 'just sync-wit' + a CI guard to catch drift. That whole pipeline was treating the symptom rather than the cause. Both bindgen macros (wasmtime::component::bindgen! and wit_bindgen::generate!) accept 'path' as an array of dirs, each holding one package. Listing both packages explicitly resolves the cross-package reference natively with no vendored copy. Changes: - crates/nexum-engine/src/main.rs: bindgen path is now ["../../wit/nexum-host", "../../wit/shepherd-cow"]; world fully qualified as "shepherd:cow/shepherd". - modules/example/src/lib.rs: world fully qualified as "nexum:host/event-module". Path stays single (the example doesn't import shepherd:cow). - wit/shepherd-cow/deps/ deleted entirely (12 files). - justfile: sync-wit recipe removed; build-runtime renamed to build-engine (clearer; matches the engine vs host vocabulary established earlier); check + run no longer depend on sync-wit. - .github/workflows/ci.yml: wit-deps-sync job removed (nothing to sync, nothing to drift). - docs/migration/0.1-to-0.2.md: checklist item about vendored deps rewritten to point at the new bindgen pattern. Result: 1 source of truth per WIT package, zero ceremony to keep them aligned, ~360 fewer lines of vendored bytes in the repo.
Updates the requirements on wasmtime to permit the latest version.
Release notes
Sourced from wasmtime's releases.
... (truncated)
Changelog
Sourced from wasmtime's changelog.
... (truncated)
Commits
377cd91Release Wasmtime 45.0.0 (#13416)1eb2c19release-45.0.0: Fix wasmtime-wasi path_open(TRUNCATE) bypass of FilePerms::WR...49b2aa8Add release notes (#13328)52e14e8Fix the version bump workflow (#13275)8bdb373gc_fuzz: Struct fields (#13101)e3e5aa6Refactor how inlining is configured in Wasmtime (#13250)85ae784Fix slow WASI stdin reads by passing size hint to worker thread (#13256)57ed7dbs390x: correct off-by-one error in rotate then select insts (#13262)244c196Warn about CLI usage of wasi-common or wasi-threads (#13264)348a8dcType-guard bitop optimization rules (#13229) (#13237)Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)