Re: backtrace_on_internal_error

Hi,

On 2023-12-08 19:39:20 -0500, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > On 2023-12-08 17:29:45 -0500, Tom Lane wrote:
> >> Agreed. I think we want to do that after the initial handshake,
> >> too, so maybe as attached.
>
> > I was wondering about that too. But if we do so, why not also do it for
> > writes?
>
> Writes don't act that way, do they? EOF on a pipe gives you an error,
> not silently reporting that zero bytes were written and leaving you
> to retry indefinitely.

Err, yes. /me looks for a brown paper bag.

> What I was wondering about was if we needed similar changes on the
> libpq side, but it's still about reads not writes.

Perhaps. It's probably harder to reach in practice. But there seems little
reason to have a plausible codepath emitting "SSL SYSCALL error: Success", so
instead mapping errno == 0 to "EOF detected" pgtls_read() and
open_client_SSL() makes sense to me.

I wish there were an easy userspace solution to simulating TCP connection
failures. I know how to do it with iptables et al, but that's not great for
automated testing in PG...

Greetings,

Andres Freund