Skip to content

v0.62.0

Choose a tag to compare

@ttraenkler ttraenkler released this 19 Jul 00:40
566f2df

v0.62.0

The standalone-conformance breakthrough release: pure-Wasm (no-JS-runtime) test262 conformance leapt from roughly 10% to 56.1%, and the test262 harness was reset to the real, rigorous upstream harness so the numbers became trustworthy.

test262 conformance

  • Standalone (pure Wasm, no JS runtime): ~10% → 56.1% — the headline. A universal harness-prelude / shim recovery (#3369) unblocked the large majority of standalone tests that previously couldn't even assemble.
  • Host (JS-host lane): 64.5%.

Harness rigor & correctness

  • oracle-v8 harness reset: the real upstream test262 harness (assert.js / propertyHelper.js / sta.js) is now authoritative instead of a thin stub — assertions genuinely fire, converting prior "vacuous passes" into honest results. Conformance figures from this release forward are trustworthy.
  • Harness compile-time speedup (#3374): the rigorous prelude exposed a quadratic compile-cost blowup on the large per-test assemblies; fixed for a 2.6–3.8× faster per-test compile with byte-identical output.

CI / infrastructure

  • test262 shard rebalance — re-weighted the CI shards for the new post-harness compile timings so the sharded suite stays balanced.

Lockstep version bump to 0.62.0.