Re: Suggested improvements to INSTALL file "Short Version"

Vicki Brown writes:

> 1) The "Longer Version" makes note of a "gmake check" option for testing.
> This SHOULD be mentioned in the short version as well!

Running the regression tests is not part of the essential installation
procedure and is there not in the short version.

> 2) It should be noted here to set the shell as /bin/sh or something similar
> (e.g. bash) or else the postmaster command below will fail. Specifically, do
> not use a Csh equivalent or /sbin/nologin.

The Bourne shell is the standard Unix shell.  Sorry.

> 3) It should be noted here that the directory /usr/local/pgsql should also be
> chown'd to postgresql,

The files under /usr/local/pgsql, except for the data directory, should
*not* be owned by the postgres user.

> or the logfile created below with the postmaster
> command will not be able to be written.

If you follow the commands *exactly* then the logfile cannot possibly be
created under /usr/local/pgsql.  If you alter some of the paths then you
need to figure out yourself what you want to do.

> 4) I can't help but wonder why the
>     make install
> didn't do the adduser, create the data directory, and set ownership
> correctly.

Because make install installs files and that's it.

> The (FreeBSD ports collection version of) PostgreSQL 7.2.2 did the
> adduser for me...

The ports collection does package management, which is different from
build management.

--
Peter Eisentraut   peter_e@gmx.net