On Fri, Nov 30, 2012 at 10:13 AM, Boszormenyi Zoltan <zb(at)cybertec(dot)at> wrote:
> now that PQconninfo() was committed, I rebased the subsequent
> patches. Actual code change was only in the last patch, to
> conform to the committed PQconninfo() API.
I've applied a modified version of the "tar unification" patch to master.
It didn't apply cleanly so I ended up redoing a number of the things
manually, but the end result is fairly similar. I don't think it'll cause
anything but really trivial merge conflicts against the 04 patch, but I
didn't actually check that (e.g. it's tarCreateHeader() not
I took at look at the basebackup patch as well. Easier to get now that it's
commented :) There's quite a lot of code in there whereby it tries to
remove recovery.conf from the tar stream if it's already in there. That
seems like an ugly way to do it. I see two other options, that I think both
are better. If we do need it to be removed, we should probably pass that as
a parameter up to the walsender, and make sure the file isn't included in
the first place. But we can also leave it in there - the tar format is
perfectly happy to have multiple copies of the same file in the archive -
it'll just use whichever copy comes last.
Given that the code there is already fairly difficult to track, I think
just keeping it simpler is definitely worth doing one of those two. I would
vote for just leaving two copies of recovery.conf in there.
Comments/thoughts from others?
In response to
pgsql-hackers by date
|Next:||From: Magnus Hagander||Date: 2013-01-01 17:25:12|
|Subject: pgsql: Unify some tar functionality across different parts|
|Previous:||From: Dimitri Fontaine||Date: 2013-01-01 16:52:57|
|Subject: Re: pg_retainxlog for inclusion in 9.3?|