Control: retitle -1 apt: deal better with non-co-installable essentials Control: severity -1 minor On Thu, Apr 14, 2016 at 11:06:58AM +0100, Klaus Ethgen wrote: > Investigating (0) dash-mksh [ amd64 ] < none -> 7 > ( shells ) > Broken dash-mksh:amd64 Kollidiert mit on dash [ amd64 ] < 0.5.8-2.2 > ( shells ) > Considering dash:amd64 5212 as a solution to dash-mksh:amd64 5206 > MarkKeep dash-mksh [ amd64 ] < none -> 7 > ( shells ) FU=0 > Holding Back dash-mksh:amd64 rather than change dash:amd64 This dash-mksh package is from a third-party repository (tg's mirbsd) and is tagged as essential – as such apt is trying to install this new essential package (as everything in this repository implicitly depends on it, so it must be installed on your system to use this repository), but later decides that it isn't going to anyhow (as it collides with another essential package: dash and keeping dash is deemed more important than installing dash-mksh [and therefore removing dash]). At this point apt will have marked mksh (which is a dependency of dash-mksh) for installation already through and reverting decisions isn't easy… So we deploy a little trick: We rely on the autoremover to clean up the mess at the end by removing the unneeded packages again from the solution. That works good enough usually but in your case the Suggests makes the autoremover consider this package as needed (as it satisfies a suggests and the autoremover isn't allowed to break suggests by default). As a solution for you, you might want to change your preferences file to pin dash-mksh instead of mksh as it is the real culpit (assuming you have a deeper reason to use that repository in the first place of course). The real "solution" here is to teach apt to not try to install dash-mksh as it isn't co-installable with an already installed essential, but that deludes the meaning of essential slightly and is hard to get right as you don't want to forbid essential-transitions downright… anyhow, a minor problem at best as conflicting essentials are far from common. (well, the super real solution is to write a better resolver, which has an easier time reverting its own decisions) Best regards David Kalnischkies
Attachment:
signature.asc
Description: PGP signature