On Wed, Mar 23, 2011 at 22:15, Jonathan Nieder <jrnie...@gmail.com> wrote: > Brian Frank wrote: >> -When running dist-upgrade from stable to testing, many of the important >> files in /bin disappeared. I found that they were provided by the >> "fileutils" package in stable, but by the "coreutils" package in testing. >> Without some of these files (mv, ln, mkdir, etc), the dist-upgrade came to >> a screeching halt.
I can't imagine a circumstance in which this could have happened. The only very very very wild wild guess is something like the fixed #590438 in which the essential is removed too early, but that wouldn't give a single hint at how to fix it if it really still exists… > Eugene V. Lyubimkin wrote: >> forcemerge 261411 282278 255969 > > To be clear: is the problem that while fileutils in stable is essential, > fileutils in testing is not, and APT was using the "essential" flag from > the latest version for some reason? As discussed in many of the threads of the merged bugreports: If any version in any repository is essential then the package is essential. So for APT a package can only 'loose' its essential flag if you remove the archive in which the package is marked as essential. The reason is simple that as soon as you have a repository in your sources you could have a package installed from it which means it could depend on an essential package implicitly. Important if you have a testing system were you install packages from unstable (possible new essential packages) and maybe packages from stable (possible packages lost their essential status since then). Thats especially important in a dist-upgrade as you have a mixed system in it all the time… This approach makes mistakes for granted and isn't optimal in all cases, but given that the consequences can be a lot worser if a essential is missing than they can be if a transition package is still installed (which is more or less the only time people really see/complain about it) i don't really see why it should be changed. And so far nobody else could provide a good reason… so i don't change what isn't broken. ;) Best regards David Kalnischkies -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
