Release 0.5.7

[madonoharu]

Prepares 0.5.7, the first release to contain Ts<T> — it landed in #71 after 0.5.6 was published, so the memory-leak fix has never reached anyone on crates.io.

Both crates, in order

The tsify-macros floor moves from 0.5.5 to 0.5.7 along with the version bump. The published 0.5.6 macro predates the Clone derive on JsType, which Ts<T> relies on — so bumping only the root crate would let an existing lockfile keep resolving 0.5.6 and break &Ts<T> parameters, the form the README recommends. The workspace builds from path dependencies, so CI cannot catch that on its own.

Publishing therefore has to go macro-first: tsify-macros → wait for the index → cargo publish --dry-run on the root against the registry version → root → tag. --dry-run on the root currently fails with failed to select a version for the requirement tsify-macros = "^0.5.7", which is the floor doing its job.

Found while reviewing the release candidate

• IntoWasmAbi for &Ts<T> leaked. It cloned the JS handle, but the shim wasm-bindgen generates for a borrowed argument never releases it, so every &Ts<T> passed to an imported function dropped a table slot on the floor. wasm-bindgen's own extern types pass the handle by reference for exactly this reason; Ts<T> now does the same, and the Clone bound falls away with it.
• The docs described code that doesn't exist. Only from_wasm_abi reaches throw_str — into_wasm_abi panics — so the CHANGELOG's "both leak the same way" was wrong. The README's #[tsify(namespace)] example still showed the pre-Make enum type aliases utilize the namespace types #78 union, contradicting this release's own CHANGELOG entry, and its .d.ts example predated the current output shape. Both are now what a build actually emits.
• The e2e gate let failures through. build_all.sh had no set -e, so a failed wasm-pack build was followed by a successful popd; compare_output.sh skipped any reference whose generated counterpart was missing. Together those two turned a completely failed e2e build into a green run. A generics test also asserted SerNamedTuple where it meant DeNamedTuple.

Also

• README's install snippet said tsify = "0.5.5" while demonstrating Ts<T>, which does not exist in 0.5.5 or 0.5.6.
• Tsify::JsType now documents the contract the ABI impls already assume: it has to be a #[wasm_bindgen] extern "C" type.

🤖 Generated with Claude Code