1. The update-alternatives setting reset should be put into some "if [ "$1" = configure ]; ...; fi" test, otherwise the command would be unconditionally executed, and not desired for situations such as abort-upgrade.
2. For the same reason, the old version $2 should be checked, so that it only attempts to do this heavy-handing fix when upgrading from known-to-be-broken versions.
3. I am still not quite sure whether preinst or postinst is the correct place. I would have put it in preinst. But this is a minor thing.
Commets about the newly proposed patch by Arne:
1. The update-alternatives setting reset should be put into some "if [ "$1" = configure ]; ...; fi" test, otherwise the command would be unconditionally executed, and not desired for situations such as abort-upgrade.
2. For the same reason, the old version $2 should be checked, so that it only attempts to do this heavy-handing fix when upgrading from known-to-be-broken versions.
3. I am still not quite sure whether preinst or postinst is the correct place. I would have put it in preinst. But this is a minor thing.