Re: mcvstats serialization code is still shy of a load

On Thu, Jun 27, 2019 at 12:04:30AM -0400, Tom Lane wrote:
>Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> writes:
>> OK. Attached is a patch ditching the alignment in serialized data. I've
>> ditched the macros to access parts of serialized data, and everything
>> gets copied.
>
>I lack energy to actually read this patch right now, and I don't currently
>have an opinion about whether it's worth another catversion bump to fix
>this stuff in v12. But I did test the patch, and I can confirm it gets
>through the core regression tests on hppa (both gaur's host environment
>with gcc 3.4.6, and the OpenBSD installation with gcc 4.2.1).
>

Thanks for running it through regression tests, that alone is a very
useful piece of information for me.

As for the catversion bump - I'd probably vote to do it. Not just because
of this serialization stuff, but to fix the pg_mcv_list_items function.
It's not something I'm very enthusiastic about (kinda embarassed about it,
really), but it seems better than shipping something that we'll need to
rework in PG13.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services