On Sun, 2024-03-17 at 15:11 +1000, Anthony Berglas wrote:
> I would write something iff there is interest in posting it.
I cannot promise that the patch would be accepted; I personally think
that briefly mentioning transaction anomalies and isolation levels
would be ok.
> Probably an overview in
On Sat, 2024-03-16 at 10:52 +1000, Anthony Berglas wrote:
> The code is wrong by any normal definition. We lose one of the updates.
> It is a very common error. It is a very common way to write code,
> especially if using an ORM, which is very common.
>
> [example of a lost update]
>
> The probl
On Thu, 2024-03-14 at 00:16 +, PG Doc comments form wrote:
> There really needs to be an explicit warning that the following is invalid
> in normal read committed mode:-
>
> select foo into f from bar where id=1;
> f = f + 123;
> update bar set foo = f where id =1;
> commit;
>
> This is a ver
