Comment 11 for bug 95422

Actually, looking at this in more detail, all the logs here (and in #94878) apart from mine show a problem reading xxx/users for an ext2/3 partition, whereas mine is a problem reading <user>/items for an NTFS (Win2K) partition.

The appearance to the user is the same (install hung in m-a), though, and the traceback is very similar. If there are likely to be many different problems that cause this sort of failure, perhaps these exceptions should be caught and the user informed that there was problem? They could then choose to skip migration which would at least allow the install to proceed. Maybe even offer to call apport to report the problem?