Re: Debian for desktop - gnome in usnstable/experimantal more stable than in testing ?
ma, 2006-01-02 kello 22:16 -0500, Benjamin Mesing kirjoitti:
> I admit I was imprecise, often it are conflicts (usually through library
> stuff) that prevent packages from being installable when you have
> certain other installed, even though you would want both.
> But as mentioned I am only repeating the experiences of other users, I
> never went with testing.
It is best not to repeat rumors, or we'll never shut up.
> Btw. can a package libterminal version 10 propagate to testing if
> package myterm in testing depends on libterminal (<= 9)? Otherwise this
> _could_ break the dependencies.
The testing distribution is all about not adding packages if it breaks
other packages. The determination of "breaks other packages" is done
primarily by checking dependencies. As long as dependencies are correct,
things won't break when packages in testing are updated or new packages
are added, or packages are removed. (Except in exception circumstances,
of course. Exceptions are not routine, though.)
See http://www.debian.org/devel/testing for more detailed information.
> But for example I can speak for my package (packagesearch) which broke,
> when xterm changed how it handles command line arguments. Of course I
> didn't knew this before, so my package depended on "xterm" (instead of
> xterm<=x.y.z). After xterm was changed, it could propagate to testing,
> breaking my package.
That was a case where dependencies were insufficiently correct.
> However due to the QT library transition my package
> which I fixed in unstable at once has not entered testing yet...
Propagation of packages into testing sometimes does sometimes take a
long time, precisely because testing has been designed to avoid random
breakage when packages are updated. Working on getting the Qt library
transition over as quickly as possible is probably the best way to
shorten the delay in this particular instance.
Teaching: the proof is in the doing.