The current code allocates from the heap after already allocating from the stack a suitably sized array. This is inefficient but also unnecessary. Allocating Vecs in library code, as a general principal, should be avoided if at all possible, and, if absolutely necessary, it would be nice to be able to capture failed allocation. It's a long standing source of problems in long-running processes.
The current code allocates from the heap after already allocating from the stack a suitably sized array. This is inefficient but also unnecessary. Allocating Vecs in library code, as a general principal, should be avoided if at all possible, and, if absolutely necessary, it would be nice to be able to capture failed allocation. It's a long standing source of problems in long-running processes.