Hi Luca, I see what you try to achieve. I've been working on some initial patches for these problems, but won't have time to get anything finished for 0.45.
The problem is something like * you want to test new.deb which is part of new.changes * new.deb is not in that archive, yet (but some old version is) * new.deb depends on more packages from new.changes * but you don't want to test all from new.changes at the same time So you provide a new package source that contains all the packages from new.changes, but that repository is only to be used when new.deb is to be installed. I'll use the codename 'target' for this partial repository (and the repository I'll call target-repository): with this you can do install tests: sid+target wheezy+target upgrade tests (that's where your "problem" started): sid -> target wheezy -> target distupgrade tests: wheezy -> sid -> target wheezy -> sid+target (skipping the installation of the version in sid) squeeze -> wheezy -> sid -> target squeeze -> wheezy -> sid+target squeeze -> sid+target Having multiple versions of the same package available at the same time is difficult for apt and piuparts - you might work around this with careful pinning and custom scripts that change the pin priorities during the test to switch the version apt will select. Check the readme, there are several environment variables that allow a script to discover when it is being run. Andreas -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
