Fix pgbench timestamp bugs.
authorThomas Munro <tmunro@postgresql.org>
Sun, 11 Jul 2021 07:50:55 +0000 (19:50 +1200)
committerThomas Munro <tmunro@postgresql.org>
Sun, 11 Jul 2021 08:08:02 +0000 (20:08 +1200)
commit0e39a608ed5545cc6b9d538ac937c3c1ee8cdc36
treee1e276196fdc79556b64481b99bbbbeb08179250
parentdd0e37cc1598050ec38fa289908487d4f5c96dca
Fix pgbench timestamp bugs.

Commit 547f04e changed pgbench to use microsecond accounting, but
introduced a couple of logging and aggregation bugs:

1.  We print Unix epoch timestamps so that you can correlate them with
other logs, but these were inadvertently changed to use a
system-dependent reference epoch.  Compute Unix timestamps, and begin
aggregation intervals on the boundaries of whole Unix epoch seconds, as
before.

2.  The user-supplied aggregation interval needed to be scaled.

Back-patch to 14.

Author: Fabien COELHO <coelho@cri.ensmp.fr>
Author: Yugo NAGATA <nagata@sraoss.co.jp>
Author: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reported-by: YoungHwan Joo <rulyox@gmail.com>
Reported-by: Gregory Smith <gregsmithpgsql@gmail.com>
Discussion: http://postgr.es/m/CAF7igB1r6wRfSCEAB-iZBKxkowWY6%2BdFF2jObSdd9%2BiVK%2BvHJg%40mail.gmail.com
Discussion: http://postgr.es/m/CAHLJuCW_8Vpcr0%3Dt6O_gozrg3wXXWXZXDioYJd3NhvKriqgpfQ%40mail.gmail.com
src/bin/pgbench/pgbench.c