Re: Why are new package versions depending on libc6 in unstable?
On Wed, Nov 20, 2002 at 09:19:17AM -0600, Steve Greenland wrote:
On 19-Nov-02, 19:20 (CST), Michael Stone <firstname.lastname@example.org> wrote:
By having developers test & use unstable.
Which is it? Do you want me to build and use unstable packages, or do
you want me to use libraries from testing?
I want people who choose to (knowing the risks) to use unstable, but I
want buildd's as much as possible to use testing.
I can make a similar assertion of the sort "how do we get testing
tested if the packages don't actually *use* testing?"
Does not compute. The packages in testing *do* use testing. They all
came from unstable.
They don't use testing, as an internally consistent whole. The packages
in testing were compiled with *some version* of their build deps from
unstable, possibly a horribly buggy or superseded one. Two packages in
unstable that depend on the same dev lib might have been compiled
against different versions from unstable, and might not compile at all
against a later version in testing. We currently work around this sort
of problem in admirably baroque ways (e.g., an email to d-d-a saying
"everyone linking to libfoo upload a new version to force a recompile
now.") This is the current analogue to the "how would we handle new
libraries" question (the answer to which might be to upload a compat
lib, which would have the added advantage of easing certain partial
upgrade problems which crop up now and then.)