On Tue, Jun 19, 2018 at 02:49:08PM -0400, Bruce Momjian wrote:
> I don't think we realize there was a behavioral change here. I think we
> were just trying to fix the case where the sequence maximum didn't match
> the serial maximum. I am not sure if it is worth documenting it at this
> point though.
Yeah, I agree that it is not worth documenting it. I don't recall
reviewing the full patch related to identity columns, but I surely
looked at patches which fixed post-commit bugs, and the new behavior is
as a whole more consistent as sequences created with serial map to the
real bound values associated with the underlying column type, and
bigserial does the same:
=# create table test (id bigserial primary key) ;
CREATE TABLE
=# select sequencename, max_value from pg_sequences;
sequencename | max_value
--------------+---------------------
test_id_seq | 9223372036854775807
(1 row)
--
Michael