On Thu, May 7, 2020 at 2:35 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Without a concrete example it's hard to say, but maybe the issue is that
> v12 is more aggressive about parallelizing restores --- see 548e50976.
Thanks Tom, I will review those changes tonight. In the meantime, to
reproduce, I run the following:
- createdb test
- create some tables, I used a simple loop to create 10 empty ones.
- pg_dump -v -Fc test | tee test.dump > /dev/null (so it's through a
pipe, same as if the file is streamed through curl which was the
original case)
- pg_restore -j4 --verbose --clean --no-acl --no-owner -d test test.dump
That should fail consistently. pg_restore_12 and pg_dump_12. Same
passes in if I run in earlier versions.
