I'm testing Cursor against other AI coding assistants. The main feature I'm evaluating is the "auto fix" for linting errors.
My process: I run a linter (ESLint) on a known problematic code snippet. When errors appear, I trigger Cursor's auto fix. My results so far are inconsistent. About half the time it resolves the issue correctly. The other half, it either makes a trivial formatting change that doesn't address the core error, or it introduces a new syntax error.
Example: a "no-unused-vars" error. Auto fix sometimes removes the variable but breaks a downstream function. It should suggest a proper fix or ask for context.
Is there a specific way to configure this? Are you supposed to highlight the exact error line first? I'm comparing pricing and reliability, and a 50% success rate on a core feature is concerning for the license cost. What's your experience with this specific function?