Coverage now skips V8 code cache during collection (
2200b3e)
Deno disables the V8 code cache whenever coverage is being collected, preventing cached modules from reporting inflated top-level coverage. This fixes a subtle but important correctness issue in coverage output and adds regression tests for run, test, and coverage flows.
mock.reset() now restores mock.method() implementations (
6adda14)
mock.reset() previously cleared call history but left method mocks installed; it now restores mocked methods as well. That brings Deno's Node polyfill behavior in line with Node's MockTracker.reset() and closes a surprising gap in test cleanup semantics.
N-API finalizer tracking switched to ordered map (
db5ac7b)
Finalizers are now tracked in a BTreeMap instead of a Vec, making registration/removal more direct and avoiding linear scans when deregistering by ID. This is a small performance and maintenance improvement in the N-API finalizer path.
Dependency patch for vulnerable rand 0.9.x (
b157cd2)
rand was bumped from 0.9.2 to 0.9.3 in the lockfile to pick up a security fix for an unsound ThreadRng aliasing issue. The update reaches Deno through hickory-resolver and opentelemetry_sdk, which are linked into the runtime binary.
Other misc changes
- Test registry update: node-gyp 12.1.0 for
@denotest/node-addonand transitive npm fixtures (1 commit) - Coverage/spec output updates for the new code-cache regression test