Marco Neumann
Marco Neumann
The proposed API looks perfect. 👍
Will apply the recommendations next week.
@abonander any update on this?
I don't think setting the cache capacity to 0 works in all cases. If you run queries w/o any transaction and the cache size is 0, SQLx will still prepare...
Another approach would be to implement proper prepared statement support in pgbouncer since sqlx isn't the only client suffering from this limitation, see https://github.com/pgbouncer/pgbouncer/issues/695
> To keep `execution_time` while fixing this issue, I would do the insert into `_sqlx_migrations` inside the transaction with `execution_time` set to 0, then commit the transaction and update `execution_time`...
I was on `0.19.1`. I've just compiled+installed the `gnome-40` branch at `0952054d35e9c9c7073ff8918b82b849b6eccc28`. I'll let you know if I still observe the issue over a day or two.
You're right :sweat_smile: Re-compiled and installed `d00e5991dacf244fa5af2ce84edc3b789bd43fc8` :+1:
With `29e2a8413ca5fa1fce4b41de0caf6d0333264db9` it's still broken: ```text Apr 22 11:11:42 mneumann-x1 gnome-shell[1128]: Object St.Widget (0x561f40c07470), has been already deallocated — impossible to connect to any signal on it. This might be...