Re: Patch to implement pg_current_logfile() function

On Fri, Jan 13, 2017 at 5:48 PM, Gilles Darold <gilles(dot)darold(at)dalibo(dot)com> wrote:
> Le 13/01/2017 à 05:26, Michael Paquier a écrit :
>> Surely the temporary file of current_logfiles should not be included
>> in base backups (look at basebackup.c). Documentation needs to reflect
>> that as well. Also, should we have current_logfiles in a base backup?
>> I would think no.
> Done but can't find any documentation about the file exclusion, do you
> have a pointer?

protocol.sgml, in the protocol-replication part. You could search for
the paragraph that contains postmaster.opts. There is no real harm in
including current_logfiles in base backups, but that's really in the
same bag as postmaster.opts or postmaster.pid, particularly if the log
file name has a timestamp.

>> pg_current_logfile() can be run by *any* user. We had better revoke
>> its access in system_views.sql!
> Why? There is no special information returned by this function unless
> the path but it can be retrieve using show log_directory.

log_directory could be an absolute path, and we surely don't want to
let normal users have a look at it. For example, "show log_directory"
can only be seen by superusers. As Stephen Frost is for a couple of
months (years?) on a holly war path against super-user checks in
system functions, I think that revoking the access to this function is
the best thing to do. It is as well easier to restrict first and
relax, the reverse is harder to justify from a compatibility point of
view.
--
Michael