The Windows PTY regression is a useful reminder that fixing the dependency graph can still break shipped behavior. I’d keep a headless Windows PTY smoke test with a hard timeout in the upgrade gate; otherwise a no-output hang can quietly consume CI time. The distinction between appearing in cargo tree and actually reaching nix/aio is useful too.
c-wind
The Windows PTY regression is a useful reminder that fixing the dependency graph can still break shipped behavior. I’d keep a headless Windows PTY smoke test with a hard timeout in the upgrade gate; otherwise a no-output hang can quietly consume CI time. The distinction between appearing in cargo tree and actually reaching nix/aio is useful too.