[rdiff-backup-users] Re: Bug report: Symbolic link replacing directory

[ruttmannn]

> There have been other threads about problems with special files
> (like symlinks).

Okay, I do not read the mailing list regularly. On the other hand: The
feature list on the homepage does clearly say, that symlinks are
supported, which they definetly are, but I couldn`t find any bug list on
the homepage. Isn´t there any?

I consider this bug as severe, as quite a lot of files can be missing
in the backup set (and were missing). I`m currently fixing that by
deleting those files and the metadata of the current backup, so the
next backup run should recreate them synchron to the source.

> Even so far as deleted _source_ trees because of symlink
> handling.

Are you sure? This would be worse than severe. As long as my source
data is untouched I always can run diffs between source and backup and
find rdiff errors, like that one, and can fix it. Not very nice, but
kinda save.

But an error condition like that is untrackable without a
backup... Can you give me a pointer to that kind of error, I really
want to know what I am dealing with here. Maybe I have to drop
rdiff-backup, because it`s to risky to use.

> What I'd like to see, is an option to have special files exist _only_ in
> metadata, not on disk. This would enable backing up to file systems
> without notion of ownership and file types, like vfat.

Would be nice, agreed. But shouldn`t it first work correctly on unix
filesystems?

cu
John