When we test concurrent transactions, one create index and one update a tuple, the execution uses codegen when we manually add sleep, but uses executor when we don't, it's really weird because the queries we use are exactly the same. And when it uses executor, bugs cause segmentation fault when accessing bw_tree index in certain ways.
When we test concurrent transactions, one create index and one update a tuple, the execution uses codegen when we manually add sleep, but uses executor when we don't, it's really weird because the queries we use are exactly the same. And when it uses executor, bugs cause segmentation fault when accessing bw_tree index in certain ways.