On Thu, Apr 06, 2006 at 08:10:14AM -0300, Henrique de Moraes Holschuh wrote: > On Thu, 06 Apr 2006, Jeroen van Wolffelaar wrote: > > net-snmp (providing libsnmp5) was forced in together with 125 other > > source packages the hard way, to resolve a huge transition with lots of > > libraries tied to eachother. > > > > Temporary uninstallability of some less important packages was a > > casuality of this, for the greater good of getting this > > I have never disputed it. I am complaining that the packages this > transition broke were *not* removed as part of the transition. And > "temporary" is *not* true. The uninstallability was *permanent* unless > fixed by a new upload of the broken packages.
The intent was temporary, and well, it ended up temporary, even though the way it got resolved was different. > Yes, that's correct. I am complaining that the transition broke hplip-base > and didn't even adopt a proper procedure of *removing* hplip-base from > testing. > > > Anyway, currently hplib-base isn't in the archive at all. > > No, I got rid of it for other reasons, and this has nothing to do with this > bug anyway. > > > Not a bug, so closing. > > Britney is to safeguard the testing archive's inner coherence. If it is > told to put 125 packages in, that's not a problem. If that transition > breaks a number of other packages, that isn't a problem either *as long as* > britney goes on and remove those packages that broke with the transition. That was not possible here if I remember correctly, either unfeasable to track down, or would involve removing way too much. > How is that not a bug?! I am going to reopen this unless you can give me a > proper technical reasopn for why testing should be broken by such > transitions with permanently uninstalable packages (which have to be *fixed* > by a new upload) instead of removing them. Maybe removing here was an option, but it probably wasn't for all packages. Anyway, you seem to be missing that this all was happening by manual intervention. The priority was to get the transition done, as good as possible. Getting so many package aligned with very elaborate interdependencies is very difficult. By putting these in, some packages got broken. It was exactly what the release managers requested from britney: "Put these packages in no matter what, even if it breaks other packages". Britney is a tool, and at the release manager's descretion, can be ordered to actually break things. This is a feature, not a bug. Whether the choice to do so was a wise one from the release managers, is a different question: I think it was, benefits of many vs disadvantages for a small number of other packages, but even if you disagree with this decision, it was a RM judgement call, and not a technical fault with any of the tools involved. Testing is still testing, and like unstable, might break from time to time, just less often. --Jeroen -- Jeroen van Wolffelaar [EMAIL PROTECTED] (also for Jabber & MSN; ICQ: 33944357) http://Jeroen.A-Eskwadraat.nl -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
