Re: Why are new package versions depending on libc6 in unstable?
On Tue, Nov 19, 2002 at 06:18:56PM -0600, Steve Greenland wrote:
> The whole point of the unstable->testing->stable system to get a stable,
> tested, consistent, releasable OS.
And how does arbitrarily depending on the libc in unstable help with that?
>A subgoal is to keep RC bugs out of
> testing. How do get that if the packages don't actually *use* unstable?
It was said before that the admin has to pick and choose which packages from
unstable are installed, that's much like a manual testing dist AFAICS.
> Yes, it's frustrating to be running testing and have to wait for the
> latest and greatest to make their way in from unstable. But hurrying
> that along by cheating just make testing more like unstable, which
> doesn't benefit the significant number of people who run testing as a
> "newer, not quite as stable as stable but pretty close" distribution.
>
No, we're not asking for anyone to hurry anything. And as for keeping
testing close to a stable state, immagine what is going to happen once
unstable's libc6 DOES finally have 10 days without any RC bugs. You'll have
libc6 2.3.x as well as a bazillion other packages that depended on it going
into testing all of the sudden. How does that keep testing anywhere near
semi-stable? In effect it makes both unstable and testing more volitile.
Now, immagine that all of those packages were NOT depending on the libc6 in
unstable. They would go in unstable, get tested by anyone useing unstable,
and hopefully find as many RC bugs as possible in the 10 days that the
package is delayed from going into testing. This happens for the several
thousand packages that depend on libc6 INDEPENDANTLY. So now when libc6
2.3.x goes into testing, there isn't a huge number of new packages that
ARTIFICALLY depended on it.
Wheather the buildds run testing or unstable, by depending on libraries in
unstable the entire unstable->testing->stable system is being hindered, not
helped.
Reply to: