Re: apt-cacher as package rollback buffer
On Mon, Feb 15, 2010 at 12:32:10AM -0800, freeman wrote:
> I installed apt-cacher because I wanted a way to reliably roll back one
> version of a package in testing. This way, if a new version is buggy, the
> way the recent xserver-xorg/mesa was for me, and I let it slip by
> apt-listbugs, I may possibly choose the old version from apt-cacher rather
> than nearly total my system messing with it, as I did.
>
I've used apt-cacher, and it worked well. However, I recommend
apt-cacher-ng (which I'm using now). I have Ubuntu and Debian machines
on my LAN, and apt-cacher did not keep the distros separate. There were
several cases where different packages from Ubuntu and Debian had the
same name and version number, and this caused problems with apt-cacher.
apt-cacher-ng solved this problem.
> It will be necessary for me to manually run the cleaning script on occasion
> to avoid random deletion of old versions. For this reason I first chose
> apt-proxy. It has a nice config allowing specification of an exact integer
> for the number or version to retain for each release. Couldn't ask for more
> than that.
>
> However, I got python errors and the hour was late.
>
> Has anyone else had success with such a plan?
>
> Is pinning really necessary or can I get by with aptitude and my apt.conf
> file:
>
> APT::Default-Release "testing";
>
This effectively pins all not-installed packages from testing at 990
(according to man apt_preferences). Are you running a mixed system? If
not, then this shouldn't be necessary, but it won't hurt.
If you are running the testing distro, and all you want to do is control
the version of a specific package, then you can either pin it (apt
preferences file) or put it on hold (using aptitude).
-Rob
Reply to: