severity 389402 important
tags 389402 pending
thanks

Hi Arnaud,

First of all, many thanks for your bug report. To my mind, it is
actually an important bug, because it touches the upgrade path Debian
and I are so proud of. :)

On Mon, Sep 25, 2006 at 04:51:26PM +0200, Arnaud Guiton wrote:
> Yesterday, I upgraded ircd-hybrid (1:7.0.3-3.1 -> 1:7.2.2-2) and found
> that one question asked by debconf was confusing.
> 
> I answered "no" to
> -------------------------------------------------------------------
> [...]
> You may choose whether or not you want to restart the ircd-hybrid
> daemon every time you install a new version of this package.
> [...]
> Restart ircd-hybrid on each upgrade?
> -------------------------------------------------------------------
> because I wanted to restart the server later, to avoid disconnecting the
> users.
> 
> Instead, it was stopped automatically and never restarted: I had to
> manually do that after the upgrade. Because I have a slow machine and it
> was a large upgrade (and also because I did not expect this), this lead
> to a pretty long downtime for the irc server.

This is indeed the case and this is fully reproducible.

> Maybe it's only a lack of understanding on my side, but is it normal
> that the server was stopped ?

You did not misunderstand anything. :) Yes, it is normal from the
way I did the maintainer scripts. You upgraded from a version which
did not support this debconf feature I introduced in 1:7.2.2-1. Thus,
the server is stopped by the old (1:7.0.3-3.1) prerm script and the
new (1:7.2.2-2) postinst script does not start it as you requested.

This did not upset me at first when I tested the upgrade path, but
now I see how confused you may have been.

> If so, shouldn't a debconf note clearly state it ?

No, debconf notes are deprecated to my discontent, but I cannot do
anything against it. :)

Anyway, the fix here concerns the upgrade path. The config script must
test if we come from a version strictly below 1:7.2.2-1. If this is the
case, then we must not ask the question. On the contrary, if we come
from a version above or equal to 1:7.2.2-1, we must ask the question.

It is nearly impossible to do otherwise. We cannot control the prerm
script from the old package which will forcefully stop the deamon.

Expect it to be fixed in 1:7.2.2-3. Well, I know it is too late
for you, but others will benefit of it for the sarge-to-etch
dist-upgrade... ;)

Cheers,
-- 
 .''`.   Aurélien GÉRÔME
: :'  :
`. `'`   Free Software Developer
  `-     Unix Sys & Net Admin

Attachment: signature.asc
Description: Digital signature

Reply via email to