On Wed, Apr 7, 2021 at 07:01:25PM -0400, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > Uh, I think your patch missed a few things. First, you use "%zd"
> > (size_t) for the printf string, but calls to pgstat_get_my_queryid() in
> > src/backend/utils/error/elog.c used "%ld". Which is correct? I see
> > pgstat_get_my_queryid() as returning uint64, but I didn't think a uint64
> > fits in a BIGINT SQL column.
>
> Neither is correct. Project standard these days for printing [u]int64
> is to write "%lld" or "%llu", with an explicit (long long) cast on
> the printf argument.
Yep, got it. The attached patch fixes all the calls to use %lld, and
adds casts. In implementing cvslog, I noticed that internally we pass
the hash as uint64, but output as int64, which I think is a requirement
for how pg_stat_statements has output it, and the use of bigint. Is
that OK?
I am also confused about the inconsistency of calling the GUC
compute_query_id (with underscore), but pg_stat_activity.queryid. If we
make it pg_stat_activity.query_id, it doesn't match most of the other
*id columsns in the table, leader_pid, usesysid, backend_xid. Is that
OK?I know I suggested pg_stat_activity.query_id, but maybe I was wrong.
--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com
If only the physical world exists, free will is an illusion.
