[arch-general] dovecot.conf moved to dovecot.pacsave - why?

[David C. Rankin]

On 02/27/2012 07:30 PM, Tom Gundersen wrote:
> On Tue, Feb 28, 2012 at 12:51 AM, David C. Rankin
> <drankinatty at suddenlinkmail.com> wrote:
>>  Why does the new dovecot package move the dovecot.conf to
>> dovecot.conf.pacsave? The same 2.0 configuration works with 2.1, why leave the
>> user without any dovecot.conf at all?
> 
> If you wonder why a packaging change was made, the first place to look
> is in the changelogs:
> <http://projects.archlinux.org/svntogit/packages.git/log/trunk?h=packages/dovecot>.
> 
>>From there you should hopefully find your answers :-)
> 
> HTH,
> 
> Tom
> 

Tom,

  Thanks, that does help, but that just leave me with more questions. I'm might
be reading it wrong, but what I get is:

<quote>
upgpkg: dovecot 2.1.1-3
don't install sample config files to /etc direclty, they break dovecot start way
too often - FS#20809; add mkcert helper script
</quote>

  That's fine and I agree. But, if we are not going to install the sample
config, why does that mean we have to move the existing dovecot.conf to
dovecot.conf.pacsave and leave the user without any config on existing installs?

  Maybe it's just a pacman limitation not knowing whether it is upgrading or
installing for the first time, but isn't there a way in an install script to
simply do a

[[ -e /etc/dovecot/dovecot.conf ]] && "don't touch it"

or

[[ -e /etc/dovecot/dovecot.conf ]] && mv /etc/dovecot/dovecot.conf \
/etc/dovecot/local.conf

  Like I said earlier, this isn't a huge issue, but I was just looking for some
better way to handle upgrades by keeping existing configs when they will still
work and nuking them when they will fail cause a fail to start.

  I guess 'one size doesn't fit all' and there is really know way for pacman to
know whether the existing config will still work or not. Even still, leaving the
existing config in such situations still seems like the better policy.


-- 
David C. Rankin, J.D.,P.E.