> Also, the Q_free function should not free q->t.
Well, that's because I thought it would own it. But if in Rust it is not implied that the Q owns the string, then I fail to see where the error is? Both l and q will be deallocated at the end of the main function. Depending on the order of the deallocation maybe q exists for a theoretical time while pointing to already deallocated memory. But both get deallocated at the same point, so q does not outlive the existence of l, so where is the error?