Hi Helmut, > > I only see three ways to get around this: > > * debian-reference-de Pre-Depends debian-reference-common (>= ...) to > > ensure the correct unpack order. > > * debian-reference-de Conflicts debian-reference-common (<< ...) to > > ensure the correct unpack order. > > * debian-reference-de takes care of removing debian-reference-common's > > symbolic link in its own preinst.
I wanted to follow your third suggestion. I guess I failed. I will think it again. Since you seem to be very knowlegeable, I have a question: Isn't the APT system smart enough to run all preinst scripts of all downloaded packages? Regards, Osamu On Sun, 2025-04-27 at 10:55 +0200, Helmut Grohne wrote: > Control: severity -1 serious > > On Thu, Apr 17, 2025 at 10:41:08PM +0900, Osamu Aoki wrote: > > As I assessed, since this bug hits user using testing only, this is not > > serious > > bug per Policy. Excuse me for my pedantic reasoning. > > As I reproduced the problem in bookworm to trixie upgrades with 2.126, > the reasoning for downgrading to important no longer applies and I am > upgrading it back to serious. > > > Rationale: There is no error seen by the official testing of Debian package > > transition checks. > > I asked the release team how to best integrate conflict checking into > testing migration checks and the answer was to use automated rc bugs. > Therefore, please consider this rc bug an official testing transition > check. It would also be possible to feed conflict checking directly into > britney2, but we opted to use rc bugs here. > > The safe way to fix this problem is upgrading the dependencies to > Pre-Depends and given that debian-reference-$LANG has no reverse > dependencies, that has a low chance of causing further problems while at > the same time reliably fixing the problem at hand by forcing correct > ordering of the symlink transition. > > Helmut >
