Currently, flags like --expose-gc and --force-node-api-uncaught-exceptions-policy are passed to all test subprocesses uniformly (see #47). Some tests require specific flags, while others may need to run without them to properly validate behavior.
The CTS tests themselves should remain implementor-agnostic — embedding Node.js-specific runtime flags (e.g. // Flags: --expose-gc) directly in test files would couple them to a specific runtime, which defeats the purpose of a conformance test suite.
Instead, this per-test flag metadata should live in the implementor harness (e.g. under implementors/node/), mapping test files to the runtime flags they require. This would allow the harness to only pass flags to the tests that actually need them, and potentially run tests both with and without certain flags where appropriate.
Ref: #47 (comment), #47 (comment)
Currently, flags like
--expose-gcand--force-node-api-uncaught-exceptions-policyare passed to all test subprocesses uniformly (see #47). Some tests require specific flags, while others may need to run without them to properly validate behavior.The CTS tests themselves should remain implementor-agnostic — embedding Node.js-specific runtime flags (e.g.
// Flags: --expose-gc) directly in test files would couple them to a specific runtime, which defeats the purpose of a conformance test suite.Instead, this per-test flag metadata should live in the implementor harness (e.g. under
implementors/node/), mapping test files to the runtime flags they require. This would allow the harness to only pass flags to the tests that actually need them, and potentially run tests both with and without certain flags where appropriate.Ref: #47 (comment), #47 (comment)