We currently bind mount `/var/lib/postgresql/data` to a local folder
which is assumed to exist with the right permissions, owned by the user
under which the `postgres` container is run (uid 1000 or the one
configured with the `RUN_USER` environment variable).
If that's not the case (on a fresh checkout, for instance), you'll
be left staring at the following error, after the container failed
fixing permissions on existing directory /var/lib/postgresql/data ... initdb: could not change permissions of directory "/var/lib/postgresql/data": Operation not permitted
Having things not work out of the box with weird errors is annoying and
can easily lead to a non-trivial amount of wasted time (hello!).
This is because Docker creates the mount point directories as owned by
root, so when the Postgres' `initdb` runs as the `postgres` user it
fails to change the ownership of the directory.
Shipping the empty directory in the repository would work, but git does
not really understand empty directories.
Sadly, even only putting a `.gitignore` there would make `initdb` complain:
initdb: directory "/var/lib/postgresql/data" exists but is not empty
It contains a dot-prefixed/invisible file, perhaps due to it being a mount point.
Using a mount point directly as the data directory is not recommended.
Create a subdirectory under the mount point.
A solution avoiding the above issues is to ship some files to
ensure the mount point directory is created with the user permissions
(assuming the user doing the checkout is uid 1000 or `RUN_USER` is set
to the right uid) and then set `PGDATA` to point to a subdirectory of
the user-writable volume mount point, making `initdb` happy.
Signed-off-by: Emanuele Aina <firstname.lastname@example.org>