On Tue, May 15, 2012 at 12:46 AM, Adam Borowski <kilob...@angband.pl> wrote: > On Mon, May 14, 2012 at 09:33:34PM +0200, Bálint Réczey wrote: >> 2012/5/14, Marco d'Itri <m...@linux.it>: >> > Do we actually have an official debootstrap package for foreign systems? >> > We could ship a static busybox with it and solve this and other issues. >> > (I know, not all the world is a VAX, etc... But we cannot solve *every* >> > problem.) >> > >> > If the problem is "foreign systems without xz" then we will never be >> > able to use XZ for base. >> How about switching to xz in the standard installer and providing an >> alternate installer with recompressed packages using gzip/bz2? >> >> This would allow us to support most systems and save on bandwidth/mirror >> space. > > Bad idea, I'm afraid. Apt really dislikes seeing two packages with the same > version but a different file size and checksum. You'd have to make sure > that alternate installer never mixes its sources with regular packages.
Where "dislikes" means it will upgrade to the first version (defined by sources.list order) with the highest pin (hint: the status file is the last source and pinned to 100), which is the reason for overriding self-build-with-same-version packages: These are usually not in a repository and therefore the last version in the listing. And the fields defining a difference in versions are: Installed-Size, Depends, Pre-Depends, Conflicts, Breaks and Replaces. Differences in all other fields are ignored (as they are not guaranteed to be present - the status file e.g. misses the deb filesize as well as checksums for obvious reasons). So in essence: You have a good chance that gz and xz packages will have the same install size and therefore be treated as equal (assuming that xz packages will be unpacked and repacked as gzip). If packages are rebuild to pack them in gzip the install-size might differ, in that case the version from the first mentioned source will be installed. Beside maybe wasting bandwidth in that case (for the sake of same version for everyone in general) a pretty well defined behavior from my POV. Best regards David Kalnischkies -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/caaz6_fbr_8a1thg+pre-4cnuweenom9el07m+w+jw__svgc...@mail.gmail.com
