> On 24 Sep 2020, at 07:45, Hou, Zhijie <houzj.f...@cn.fujitsu.com> wrote:
> I found some tables is not dropped at the end of the sqlscript, > It does not hava any problem, but I think it's better to drop the table in > time. There is value in keeping a representative set of objects around after regression tests, as the database which is left after regression tests run is used as the input to the pg_regress tests. That being said, I don't think that was the intention for these relationns in commit 92c58fd94801, but in general we should avoid cleaning up too much to ensure that we stress pg_upgrade well enough (there is a delicate balance to keep test runtime short as well of course). cheers ./daniel
