On lun., 2011-04-18 at 18:05 +0200, Patrick Ohly wrote: > On Mo, 2011-04-18 at 17:50 +0200, Yves-Alexis Perez wrote: > > On lun., 2011-04-18 at 17:46 +0200, Patrick Ohly wrote: > > > How is that going to help a user who has Evolution and SyncEvolution > > > installed on Squeeze and then does a partial update of just Evolution > > > with "aptitude install -t testing evolution"? > > > > > > If I'm not completely mistaken, aptitude will complete leave > > > SyncEvolution alone and installed in the Squeeze version together with > > > the old libecal/ebook, thus leading to a broken system. > > > > It'll prevent upgrading the lib to 2.32 if syncevolution isn't updated > > together. > > Why? That doesn't follow for me. > > Here's the situation before the update (focusing on libecal): > * installed from Squeeze: > * syncevolution 1.0+ds1~beta2a-2, depends (>= 2.30.3) on > * libecal1.2-7 > * Testing has: > * syncevolution 1.1~foo-bar, depends (>= 2.30.3 <= 2.32) > on > * libecal1.2-8 2.32.2-4 > * Evolution 2.32.2-2, same dependency > > There's nothing that prevents installing Evolution 3.32 + libecal1.2-8 > from Testing while keeping libecal1.2-7 2.30.3 and syncevolution 1.0 > +ds1~beta2a-2 installed.
HMh yeah good point, the soname change means the depends won't work, it'll be satisfied indeed. So shlibs is not enough. > > Basically the apps (SyncEvolution or Evolution) can't and shouldn't know > whether libecal1.2-7 and libecal1.2-8 are installable in parallel. Same > for a future libecal1.2-9. That will only be known when that future > version is ready and gets packaged, at which point a "conflicts with" > should be added to the new package based on that future knowledge. No, I don't think that's a good idea. In that *specific* case, evolution is the one updating the gconf schema, aiui, so I guess a fix would be to add a Breaks: libecal/libebook to evolution 2.32. I'm still not so sure about the responsibilities in that case, sorry. Regards, -- Yves-Alexis -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
