Horton

Re: Adding facility for injection points (or probe points?) for more advanced tests

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Amul Sul <sulamul@gmail.com>
Cc: Postgres hackers <pgsql-hackers@lists.postgresql.org>
Date: 2023-10-25T04:57:18Z
Lists: pgsql-hackers 
On Wed, Oct 25, 2023 at 10:06:17AM +0530, Amul Sul wrote:
> +1 for the feature.
> 
> TWIMW, here[1] is an interesting talk from pgconf.in 2020 on the similar
> topic.
> 
> 1] https://pgconf.in/conferences/pgconfin2020/program/proposals/101

Right, this uses a shared hash table.  There is a patch from 2019 that
summarizes this presentation as well:
https://www.postgresql.org/message-id/CANXE4TdxdESX1jKw48xet-5GvBFVSq%3D4cgNeioTQff372KO45A%40mail.gmail.com

A different idea is that this patch could leverage a bgworker instead
of having a footprint in the postmaster.  FWIW, I think that my patch
is more flexible than the modes added by faultinjector.h (see 0001),
because the actions that can be taken should not be limited by the
core code: the point registered could just use what it wants as
callback, so an extension could register a custom thing as well.
--
Michael

Commits

  1. ci: Enable injection points in builds

  2. Fix two memcpy() bugs in the new injection point code

  3. Add test module injection_points

  4. Add backend support for injection points

  5. Refactor code checking for file existence