Re: pg_createsubscriber --dry-run logging concerns

On 2025-Nov-10, Peter Smith wrote:

> Hi Alvaro,
>
> Here is patch v4-0001 modified as requested:
> - dashes are removed
> - the message is a single string

Okay, thanks. I split the strings in two lines, as we customarily do
when they contain embedded newlines. I also noticed pg_resetwal uses
stdout rather than stderr and set out to change it, because I don't
think it's sensible to have one program behave one way (print to stdout)
when all others behave in another (to stderr). I wrote a commit message
and was about ready to push.

However, I then found out that the reason you used stdout instead of
stderr in pg_resetwal is that with the latter, tests fail all over the
place because of pg_resetwal -n being used for pg_upgrade internally via
popen(), and making it write to stderr results in confusing pg_upgrade
output as well as test failures. A very simple fix for this problem
would be, of course, to add " 2>/dev/null" to the popen call, but that
is not only cheating, it is also dangerous: if pg_resetwal ever finds
reason to complain, we won't get very good information because of that
redirection.

(I also don't think this line belongs in stdout, in case you're thinking
of changing it in the other direction for all other programs.)

Maybe we should add a -q,--silent mode that suppresses the "Running in
dry-run mode" line. I do wonder if this is getting too far into the
weeds for such a small thing. I won't blame you if you want to just
drop this whole idea, but I also won't stop you if you want to introduce
--silent.

--
Álvaro Herrera Breisgau, Deutschland — https://www.EnterpriseDB.com/
"No necesitamos banderas
No reconocemos fronteras" (Jorge González)