On Wed, Oct 31, 2007, Vasilis Vasaitis wrote: > I've been using unstable successfully for many years now, I don't > need the patronising thanks very much. And besides, how are you going > to have testing be relatively bug-free if you try to drive away from > unstable the users who actually care enough to report the bugs, so > that they can get fixed before the packages hit testing?
You have been using unstable successfully for many years but yet you still wonder about the very natural uninstallability issues it naturally carries. Concerning testing, as the "testing" name implies this is where testing by end users should happen; unstable is precisely there for transitions and major changes. Sure, some testing happens in unstable as well, but by people who understand what kind of disruptive changes might happen in unstable. Now I'm not saying that you wouldn't ever be able to do testing in unstable yourself, but I'm rather claiming that if you don't know about the difference between unstable and testing, then you shouldn't run unstable. > Anyway, you didn't answer my question at all. All I see is this > package being uninstallable and its maintainers pointing fingers > elsewhere instead of fixing it. Now obviously from your point of view > you're doing the right thing, so it'd be nice if you could explain why > this is so instead of trying to avoid the subject altogether. Then let me answer your question: we regularly get uninstallability issues in unstable which are a natural situation when a SONAME changes and when a package is renamed as a consequence. It should be clear to the Debian Developer introducing such a change that he should contact the release team to request rebuilds (bin NMUs) of the rdeps when that's required; reassigning is not the correct solution to the installability problem, in fact the bug tracking system is the wrong place to discuss any of this. This is why maintainers of affected packages which can do nothing with the bug reports they receive in such a situation reassign the bugs to the package introducing the change as the maintainer of this package should get the blame. You can read parts of the discussion I had with the maintainer in <http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=447023>. The only concrete action which can be done (and which I did) is to contect the release team. Since I contacted them, there were at least 4 other requests by random people (who should have checked whether this had been requested already) and many more random bugs. It was a serious tip when I told you you should use testing; it simply is made to prevent/hide these issues. -- Loïc Minier
