Re: apt-get upgrade message in sid

2011-02-14 Thread David Kalnischkies
Hi debian-user :)

 What is David Kalnischkies telling me here?

He tries to tell you that apt-get will try a minimal release override
for you according to the dependencies in the request.

Lets assume you want to install iceweasel version whatever from experimental
or backports. This includes normally the installation of a few new packages
(xulrunner-foo) and the upgrade of some (lets say libmozjs3d).
Previous APT would have installed xulrunner-foo from the other release if
it was the only release providing this package. But libmozjs3d was already
available in stable but in a lower version, so previously APT would have
favored libmozjs3d from stable which can't satisfy the dependencies so
APT happily blows up with an error message telling you that dependencies
can't be satisfied.

What APT now tries is, while choosing the version of iceweasel
based on your request, it looks at the dependencies of your requests
and checks if these can be satisfied by the current candidate of the
package and if not it tries to switch the version of this package, too.

So in the iceweasel thing above it would install libmozjs3d from
experimental, too, which is very very likely what you wanted -
as nobody wants to see an error message as a respond to a request.

It's not a new solver strategy or anything, it just tries to help a
bit by expanding the request with packages you need to switch, too.


Thats why stuff like libc6 from experimental fails tragically:
The request is expanded to libc-bin as this one is versioned.
Fine so far. The sole problem is now that stuff like libc6-i686
needs a specific version of libc6 - thats a reverse dependency in
the eyes of libc6 and reverse dependencies are not touched.

(Beside, in this very specific case libc6-i686 is also a recommends of
 libc6 so the request would work if the recommends would be versioned…)


Best regards

David Kalnischkies

P.S.: Next time, if you are talking about a specific guy, feel free to
at least cc him - feels strange to stumble across threads mentioning
your name by accident only…

P.P.S.: I am not subscribed, so please cc me in response.


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/aanlktimtwbj6jyyifg8zdka7fccgkd_n5v74ikm-m...@mail.gmail.com



Re: apt-get upgrade message in sid

2011-02-14 Thread David Kalnischkies
On Mon, Feb 14, 2011 at 10:57, Keshwarsingh Nadan k...@debian.mu wrote:
 Lets assume you want to install iceweasel version whatever from
 experimental or backports

 If we wanted to install a testing version of a program (for example), we
 would have to override the choices we make when we use apt-get, e.g.
 apt-get install [packagename]/testing,   or if necessary   apt-get -t
 testing install [packagename].

And thats is the point: APT wants to help with this feature to reduce the
times a user has to use '-t experimental' (or '-t backports') in favor of
'/experimental' as switching the whole release for every package is more
dangerous than for a very limited set of packages - but in the past it was
way 'easier' to just say '-t experimental'.
Its still 'easier' but the alternative '/experimental' became more useful…


 Note that another option would be to
 momentarily make testing the highest priority in /etc/apt/preferences, then
 override what will be installed, e.g.   apt-get install
 [packagename]/testing. Read this. At this time however, we are not using apt
 pinning.

Read this should btw remove the section about MMap running out of the
room. The issue is fixed in squeeze as the MMap grows now automatically
(on all architectures by the way, even kfreebsd).

But yes, i can sign that preferences are not for beginners - its a very
powerful feature, but as we know with great power comes great responsibility…
I would recommend to have a look at 'man apt_preferences' for details.

Using preferences in this usecases isn't a good idea, as -t does exactly the
same as a preferences entry for testing would do (with a pin-priority of 990).


Best regards

David Kalnischkies


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/AANLkTi=4qzly7x_gg4fp-v4a8jgdyxaaacv5xiwuz...@mail.gmail.com