On Tue, Mar 07, 2023 at 02:03:05PM +0100, Daniel Gustafsson wrote:
> On 7 Mar 2023, at 09:26, Daniel Gustafsson <daniel@yesql.se> wrote:
>> Right, what I meant was: can a pg_regress sql/expected test drive a psql
>> interactive prompt? Your comments suggested using password.sql so I was
>> curious if I was missing a neat trick for doing this.
Yes, I meant to rely just on password.sql to do that. I think that I
see your point now.. You are worried that the SET command changing a
GUC to-be-reported would not affect the client before \password is
done. That could be possible, I guess. ReportChangedGUCOptions() is
called before ReadyForQuery() that would tell psql that the backend is
ready to receive the next query. A trick would be to stick an extra
dummy query between the SET and \password in password.sql?
> Running interactive tests against psql adds a fair bit of complexity and isn't
> all that pleasing on the eye, but it can be cleaned up and refactored when
> https://commitfest.postgresql.org/42/4228/ is committed.
I have not looked at that, so no idea.
--
Michael
