Re: psql not responding to SIGINT upon db reconnection

Tristan Partin <tristan@neon.tech>

From: "Tristan Partin" <tristan@neon.tech>
To: "Gurjeet Singh" <gurjeet@singh.im>
Cc: <pgsql-hackers@postgresql.org>
Date: 2023-07-24T17:09:05Z
Lists: pgsql-hackers
On Mon Jul 24, 2023 at 12:00 PM CDT, Gurjeet Singh wrote:
> On Mon, Jul 24, 2023 at 9:26 AM Tristan Partin <tristan@neon.tech> wrote:
>
> > attached patch
>
> +        /*
> +         * Restore the default SIGINT behavior while within libpq.
> Otherwise, we
> +         * can never exit from polling for the database connection. Failure to
> +         * restore is non-fatal.
> +         */
> +        newact.sa_handler = SIG_DFL;
> +        rc = sigaction(SIGINT, &newact, &oldact);
>
>     There's no action taken if rc != 0. It doesn't seem right to
> continue as if everything's fine when the handler registration fails.
> At least a warning is warranted, so that the user reports such
> failures to the community.

If we fail to go back to the default handler, then we just get the 
behavior we currently have. I am not sure logging a message like "Failed 
to restore default SIGINT handler" is that useful to the user. It isn't 
actionable, and at the end of the day isn't going to affect them for the 
most part. They also aren't even aware that default handler was ever 
overridden in the first place. I'm more than happy to add a debug log if 
it is the blocker to getting this patch accepted however.

-- 
Tristan Partin
Neon (https://neon.tech)



Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →
  1. Fix old, misleading comment for PGRES_POLLING_ACTIVE.

  2. Remove reachable call to pg_unreachable().

  3. Allow SIGINT to cancel psql database reconnections.

  4. Expose PQsocketPoll via libpq

  5. Fix query cancellation handling in psql