Suppressing noise in successful check-world runs

[ redirected from a thread in pgsql-committers[1] ]

As of commit eb9812f27 you can run a manual check-world with
stdout dumped to /dev/null, and get fairly clean results:

$ time make check-world -j10 >/dev/null
NOTICE: database "regression" does not exist, skipping

real 1m43.875s
user 2m50.659s
sys 1m22.518s
$

This is a productive way to work because if you do get a failure,
make's bleating gives you enough context to see which subdirectory
to check the log files in; so you don't really need to see all the
noise that goes to stdout. (OTOH, if you don't discard stdout,
it's a mess; if you get a failure it could easily scroll off the
screen before you ever see it, leaving you with a false impression
that the test succeeded.)

However ... there is that one NOTICE, which is annoying just because
it's the only one left. That's coming from the pg_upgrade test's
invocation of "make installcheck" in the instance it's just built.
(Every other test lets pg_regress build its own temp instance,
and then pg_regress knows it needn't bother with "DROP DATABASE
regression".)

I experimented with the attached quick-hack patch to make pg_regress
suppress notices from its various initial DROP/CREATE IF [NOT] EXISTS
commands. I'm not entirely convinced whether suppressing them is
a good idea though. Perhaps some hack with effects confined to
pg_upgrade's test would be better. I don't have a good idea what
that would look like, however.

Or we could just say this isn't annoying enough to fix.

Thoughts?

regards, tom lane

[1] https://postgr.es/m/E1hSk9C-0002hH-Vp@gemulon.postgresql.org