[clue-tech] yum / apt update minimum age

[David L. Anselmi]

David L. Willson wrote:
>>> Here's my objective:  I want to deploy every package that's been 
>>> available and newest for a week.  My hope is that by doing so, I
>>> will avoid those rare times that a bugged patch gets uploaded, breaks 
>>> everything, and then a new, good patch goes up the next day.
> 
> I do, actually.  Sometime last year, I spent most of a day
> troubleshooting an inaccessible Ubuntu Samba server.  I finally did a
> full update on the system, even though it had ~just~ been updated,
> with a patch for Samba even.  Lo, there was a new patch for Samba and
> after applying it, everything began to work again.  I don't know for
> ~sure~ that the prior patch was bugged, but I suppose it was.  I've
> seen lots of double-patches go by, though.  A patch one day, then
> another patch the next.  It seems reasonable to wait a day or three
> on a patch being newest before deploying it.

Seems like a good reason not to use Ubuntu, if they're doing that to a 
public release.  Debian's unstable works just like that.  But there's a 
10 day wait to go into testing so I really doubt that testing and stable 
see double-patches.

Here's an example that's much more common for me (Debian unstable, so 
I'm *not* asking avoid double-patches).

I tried to use debmirror yesterday to mirror hardy (Ubuntu).  The same 
command line I've always used failed with an odd message.  I looked at 
the changelog--they "fixed" something related to suite/codename 
symlinks, whatever those are.  But that bug was already fixed: it 
prevented mirroring Debian experimental too.  So I did another upgrade 
and got a different, odd error.

So I filed a bug report.  It's fixed today (the maintainer is a few 
timezones ahead of me, which didn't hurt).

My experience is that by the time I find a bug it's either a) fixed 
because someone else found it first, or b) a new bug because I'm doing 
something a little different than "average".

Dave