5 Jul
2019
5 Jul
'19
2:49 p.m.
Hi, Aleksey! On Jul 05, Aleksey Midenkov wrote:
Hello Sergei!
Actually I was aware of that, but open_purge_table() looks so non-generic, so I decided to push that down. Well, I agree, it's "kind of" generic and it will be safer to move a bit higher. But why to remove DEBUG_ASSERT? update_virtual_field() relies on is_error() from inside. If error comes from outside it will return TRUE illegally. Like it happened with current bug. If there were assert in the first place it didn't took so long to debug.
Ideally I'd prefer update_virtual_field() not to rely on thd->is_error(). But changing this is clearly out of the scope of this bug. So okay, let's keep the assert. Regards, Sergei VP of MariaDB Server Engineering and security@mariadb.org