Re: BGWORKER_BYPASS_ALLOWCONN used nowhere (infra part of on-line checksum switcher)

On Sun, Apr 22, 2018 at 1:11 PM, Michael Paquier <michael(at)paquier(dot)xyz>
wrote:

> Hi all,
>
> BGWORKER_BYPASS_ALLOWCONN has been added by commit eed1ce7, which is an
> infrastructure piece to be able to enable and disable dynamically
> checksums on a cluster. The main idea is to be able to bypass
> datallowconn which allows a background worker to connect to a database
> even if the database is set to refuse connections so as its checksums
> can be calculated and updated.
>
> At the end, the dynamic switch for checksums has been reverted as of
> a228cc13, and a set of rather-used APIs have been changed for what looks
> like no reason now:
> - BackgroundWorkerInitializeConnection
> - BackgroundWorkerInitializeConnectionByOid
> - InitPostgres
> So all background workers would not be able to compile because of that.
> Would we want to drop this unused interface or keep it?
>
> Even if this is not removed, bgworker.sgml needs to be updated with the
> new definition of BackgroundWorkerInitializeConnection and
> BackgroundWorkerInitializeConnectionByOid which are missing the third
> argument "uint32 flags", as well as the description for
> BGWORKER_BYPASS_ALLOWCONN. I can personally see more reasons to revert
> that portion as well and consider it again for v12 or onwards if the
> on-line checksum switch is proposed again.
>
> Magnus, Daniel, what do you think?
>

I think this feature is definitely worth keeping, regardless. It's useful
elsewhere, in externally maintained bgwriters.

The incorrect documentation is clearly a bug of course, and I will fix.
Good spot!

--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/>
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/>