On Tue, Jul 10, 2018 at 03:31:09PM +0200, Hans wrote: > Hi folks, > > I know, I had discussed a similar question ago, but testing is still a > miracle > for me. > > Is there an automatism, why or when packages got removed from testing or ist > this always done manually by the developers? > > This time I struggled about "cqrlog", which was accepted in testing, then 3 > months later removed without a clear reason but same version added in sid? > Clear reason is here:
https://tracker.debian.org/pkg/cqrlog testing migrations excuses: 348 days old (5 needed) Updating cqrlog introduces new bugs: #867140 Piuparts tested OK - https://piuparts.debian.org/sid/source/c/cqrlog.html Not considered > However, I discovered, there is a new version available, but has broken > dependencies (just a single lib is missing!) . > > But the removal of "cqrlog" looked as it was done by some cronjob. > That is correct. When a package in testing has a release-critical bug filed against it, the package is scheduled for automatic removal after some period of time. If the maintainer responds with an upload the fixes the bug, the package is allowed to remain. If the maintainer takes no action the package is removed. This is done ensure a high quality set of packages for the next release. Recall that the purpose of the testing distribution is not to serve as a more up to date stable version for users. It is meant to be the preparation area for the next stable release. The bug that triggered the removal of cqrlog (#867140) has to do with the transition to openssl 1.1. > On the other hand, it is not understandable, why to remove a package, when > its > dependencies touches other packages. When it was prior existent, then the > dependencies of the later package should be adjusted and not the existing one > (or just remove it). > Most packages that are built against openssl 1.0 need non-trivial source code changes to work with openssl 1.1. Many packages have been removed as a result of this transition. > The logic of debian/testing is still a miracle for me, looks like changes are > done one time so, the other time so. > > Sorry, please do not feel beeing attacked, it is just the way it is looking > for me. :) > > Testing is a miracle.... > Again, the logic has to do with the intended purpose of testing. If you need stability, then run stable. If you can deal with the occasional disruption, then unstable is probably best for staying current with packages. Testing should really only be used by those are specifically working toward the development of the next Debian stable release. Regards, -Roberto -- Roberto C. Sánchez