Re: pgsql: Support "postgres -C" with runtime-computed GUCs

On Tue, Sep 21, 2021 at 12:47:24PM +0900, Michael Paquier wrote:
> This would print the parameter, while getting to stderr mostly a
> "could not start server". Would that work for you?

And.. Here is an actual patch. The first test works here on Win32,
but I don't have a CI environment with an unprivileged account. The
second is funky, as pg_ctl would return 0 causing a failure with
command_fails_like() because of the way start_postmaster() differs in
pg_ctl.

it may be possible to work around that by switching to
command_checks_all() but that would make the whole uglier with
different error handling, and I guess different reports generated, so
I would just remove it. Could you check if the first case would work
in your environment? If not, I would not mind just removing the test.
I have discovered a separate bug thanks to it, but that's not worth
breaking your case either.
--
Michael