On Thu, Aug 14, 2003 at 12:14:37PM -0400, David Z Maze wrote: > "Jacob Anawalt" <[EMAIL PROTECTED]> writes: > > The fact that I hear it said so much re-confirms to me that the labels > > testing and unstable (unless you like adventure) and possibly the > > writeup on the debian site lead the masses to expect something > > different than what they get. > > Testing is still a comparatively young concept for Debian. For the > woody release, it seemed to work fairly well. At this point, the > canonical problem we've run into looks something like this: package A > depends on package base (= 1). Base is upgraded to version 2. > Packages B through Z update and now require base (>= 2), but A > doesn't. The way the testing rules work, base can't be updated, since > that would break A's dependency, but that means that none of B through > Z can be updated either. The solutions are either "wait forever" or > "intentionally break A", and the testing czars have gone with the > latter option as of late, with the result that testing is closer to > current but not necessarily useful on its own.
There's a third solution, which is "wait until everything's ready and then give the testing scripts a hint that they can update A to Z all at once", which is used quite often. It's difficult (not impossible, but difficult) for large groups of packages, though. -- Colin Watson [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
