I'm looking at the proposed fix for this bug and have a few concerns. I hope they can be addressed before the next upload.
1. What about the possibility that /var/log and /var are different filesystems? The preinst may need some error checking to handle out-of-space conditions from mv, and the administrator needs to be warned in advance about the move (I think this calls at least for a mention in the release notes). Moreover, I'm not convinced that the currently proposed preinst script is idempotent (policy ยง6.2). How about adding a symbolic link from /var/account to /var/log/account on upgrade instead? 2. What happens if both /var/account and /var/log/account already exist? Is it really appropriate to try and clobber an existing /var/account without consulting the administrator? Again, consider the idempotency requirement. 3. I'm not sure offhand whether process accounting will be turned off while the preinst script is running. (The old prerm in buster only stops the acct service on remove, not on upgrade.) If not, there is a risk that accounting records will be written to the old location during the move. (At the same time, I'd like to keep any gap in process accounting coverage at a minimum; but this is a wish, not a requirement.)
