Jack Christensen

I looked over the code and it is an improvement to have the logic in pgconn. But as I mentioned in a previous reply, I'm still unconvinced this is the...

Here's a proof of concept of the approach I described: https://github.com/jackc/pgx/pull/1894. It doesn't prevent adding recovery the ability to recover from a deadline interrupting a network read, but I think...

> If i want to simply disable HandleCancel or define it with other functionality, that means we wont get read error and that means we dont get context cancelation error....

So the problem is that if a handler delays interrupting the query such that the query is successfully completed that the query still returns a context error even though no...

Ah, I see now. I actually don't think there should be an error. Returning the context error says the query failed because of a context cancellation. But the query did...

> Thats true on the one hand, but on the other it makes using library very inconvenient because user would have to check context after each query I guess I...

Thanks, I see what's going on. But IMO the correct solution is for `someLongMathOperationOrSomeLongExternalCall()` to accept a context and handle its own cancellation. Otherwise, a context cancellation right as the...

> Any updates here? Have you made up your mind about the approach to this issue? Do you think we could merge this pr into some experemental branch and give...

I'm conservative about exporting errors because they add to the public interface in a way that can be difficult to guarantee doesn't change. But I'm open to exporting more specific...

Can you provide a reproduction case?