Closed mikecaat closed 1 year ago
BTW, the costs changed because the number of relpages
is changed in my laptop. It seems reasonable because the commit
changes the logic of extending relation.
# One commit before postgres/postgres@d37ab37
=# SELECT relpages FROM pg_class WHERE relname = 'tpc';
relpages
----------
5
(1 row)
# postgres/postgres@d37ab37
=# SELECT relpages FROM pg_class WHERE relname = 'tpc';
relpages
----------
8
(1 row)
Okay, thanks for this one. Applied down to 11, even if the difference only showed up on HEAD.
The commit adds the EXPLAIN OPTION '(COSTS false)' for test stability, because some tests don't need to check the costs.
make installcheck
is failed after https://github.com/postgres/postgres/commit/d37ab378b6e773c278c14b9554a1ea23b355aab9 is committed because the costs is changed. (Might this only happen in my environment?)Although only ut-A is failed on my laptop, I also fix ut-T because I have found that it can cause the same problem with the following query.