'testing' packages under 'stable'
Nick Phillips <nwp@lemon-computing.com> writes:
> The other problem that I hinted at above is the fact that too many
> users feel that they have to move over to testing.
>
> Given the way that Debian develops, if a user requires a new feature
> that's only present in a testing package, they often^Wusually have
> to essentially go the whole hog and upgrade their whole system to
> testing. I don't believe that this is a desirable state of affairs
> (although with the advent of testing it's certainly a whole load
> better than it used to be, where such users were required to go all
> the way to unstable).
>
> This appears to be largely due to the way shared library
> dependencies are calculated, and the fact that most people seem to
> do development on bleeding-edge systems.
>
> Surely there must be some way that we can improve this situation -
> obviously some packages will require major upgrades (like, it really
> won't work without perl 5.6, or the latest libc or whatever), but I
> reckon a whole load would work just fine if they'd been built on a
> stable system - as shown by the number of private repositories of
> new-packages-built-for-potato that exist.
It might be sufficient to specifically document the procedure for
building a package from testing under stable, given that it often is
no more difficult than unpack, build, binary.
The downside of doing this is the user who does it has to be more much
careful about securing their system, given that AFAIK security.d.o
only carries updates for stable. i.e. the same piece of documentation
should mention precisely such caveats as this one.
--
http://www.greenend.org.uk/rjk/
Reply to: