Re: BUG #17485: Records missing from Primary Key index when doing REINDEX INDEX CONCURRENTLY

On 2022-May-31, Andrey Borodin wrote:

> As seen in CI with reverting only d9d0762:
>  v3 fails only on MacOS [0,1]
>  v4 fails on Linux and FreeBSD [2,3]

> [0] https://github.com/x4m/postgres_g/runs/6665396918
> [1] https://cirrus-ci.com/task/4580203115577344?logs=test_world#L1571
> [2] https://github.com/x4m/postgres_g/runs/6675153300
> [3] https://cirrus-ci.com/task/6540481319403520?logs=test_world#L842

Hmm ... Exactly what code is being tested here?  I see in one of these
backtraces that there is a reference to a function called
heap_prune_from_root, but I can't find that in the Postgres sources.

At this point, I'm clear that d9d076222f5b needs to be reverted because
of the interactions with HOT prune, but it's not at all clear to me what
is the relationship of the other two commits to bugs.  I am going to
revert that one now, so that we can continue to diagnose any lingering
problem that may be related to the other two commits -- but perhaps it
is something else.

-- 
Álvaro Herrera               48°01'N 7°57'E  —  https://www.EnterpriseDB.com/