Austin Bonander

Yeah, this is a backwards compatible change however, so if anyone wants to put in a PR for it we'd gladly have it.

Closed by #1953

I do agree, it seems strange for the insert into `_sqlx_migrations` to happen outside of the transaction. I believe it's so that `execution_time` includes the transaction commit (which is where...

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` afterward.

Postgres doesn't support nested transactions. We simulate them using [savepoints](https://www.postgresql.org/docs/current/sql-savepoint.html), which wouldn't work for that. There's [`PREAPRE TRANSACTION`](https://www.postgresql.org/docs/current/sql-prepare-transaction.html) which does _most_ of the work of committing a transaction without fully...

@zzhengzhuo I'm really sorry for leaving this open for so long, when PRs get buried on the second page of results it gets easy to forget about them. Do you...

@zzhengzhuo sorry for the delay again, the `corrupt message` error can happen when trying to use RusTLS to connect to an older version of MySQL. It might be a latent...

@zzhengzhuo it's passing but now there's merge conflicts, do you have time to resolve those?

Great, glad to finally land this!

### All types OID sources: https://github.com/postgres/postgres/blob/2743d9ae4a490a7d96b5c19d50694bd101a87dc8/src/include/catalog/pg_type.dat Text encoding descriptions: https://www.postgresql.org/docs/current/datatype-geometric.html The documentation specifies which text syntax you're looking for. For robustness, the implementation should not assume any specifics about whitespace...