Re: Issues with ON CONFLICT UPDATE and REINDEX CONCURRENTLY

 Hello, Álvaro!

On Fri, Dec 12, 2025 at 11:17 AM Álvaro Herrera <alvherre@kurilemu.de> wrote:
> Another idea might be to rewrite these tests using BackgroundPsql under
> the TAP infrastructure.  That's quite a bit more tedious to write, but
> we can be more precise on detecting whether some particular error
> message was thrown or not.

I think I understood the race, currently thinking about two possible approaches:

1) extract LOOP waiting for injection point into some function like
"injection_points_await_waiter" and add it almost between each steps
to ensure it all executed by guardrails (effectively reducing
concurrency instead of making isolationtester to report data the same
way)
2) rewrite using TAP infra

What do you think about this? Which one do you prefer?

Best regards,
Mikhail.