The same failure happens if you "upgrade" from precise to precise using this same path (which requires modifying ubiquity in-place to permit this, but I just wanted to check the behaviour of precise's ubiquity). This therefore isn't a regression.
We probably need to arrange to hide the user-setup page in the upgrade-in-place case and instead reuse the user credentials from the pre-upgrade system; or else, if we were to allow entering a new password, we'd also need the old password to be entered so that we could rewrap the user's ecryptfs passphrase.
The same failure happens if you "upgrade" from precise to precise using this same path (which requires modifying ubiquity in-place to permit this, but I just wanted to check the behaviour of precise's ubiquity). This therefore isn't a regression.
We probably need to arrange to hide the user-setup page in the upgrade-in-place case and instead reuse the user credentials from the pre-upgrade system; or else, if we were to allow entering a new password, we'd also need the old password to be entered so that we could rewrap the user's ecryptfs passphrase.