On 09/04/2016 12:10 PM, Rich Shepard wrote:
> On Sun, 4 Sep 2016, Adrian Klaver wrote:
>
>> You don't it directly. That information is supplied by Postgres when you
>> do CREATE or ALTER ROLE. The -W switch just does that for the
>> superuser(postgres in your case) when you initdb a new cluster.
>
> Adrian,
>
> OK. That makes sense.
>
>> Sorry, old habits. pg_user is a version of the pg_shadow view that blanks
>> out the actual password. pg_shadow is a view over the table pg_authid,
>> where the actual information is stored now. In any case, again they are
>> not tables/views you directly modify.
>
> Good. Then I won't spend time with them.
>
> So, given my single-user situation do you think that I should ALTER ROLE
> to add my password? Adding it to ~/.pgpass did nothing positive when I
> changed auth method to md5; my attempt to open a database failed because
> that password was rejected. Strange ... to me.
Well first, if you are going to use trust as your auth method then
specifying a password is moot exercise.
Second, not sure where you are in the process, but any time you change
the pg_hba.conf file you will need to give Postgres a reload signal to
get it to recognize the changes. Again not sure how you are signalling
Postgres but if you are using pg_ctl
https://www.postgresql.org/docs/9.5/static/app-pg-ctl.html
then something like:
pg_ctl reload -D path_to_your_datadir
as OS user postgres.
Third, .pgpass should hold information that already exists in the
database system tables. It is not a mechanism for entering that
information into the database. So yes, you will need to use ALTER ROLE
to create the password inside Postgres.
>
> Rich
>
>
>
--
Adrian Klaver
adrian.klaver@aklaver.com