Re: "readlink -f foo" fails if the target of foo does not exist

[Thomas Hood]

severity 225836 wishlist
retitle 225836 Please implement "--recursive" option
thanks

On Fri, 2004-01-09 at 11:29, Jim Meyering wrote:
> The problem with the previous implementation is that its behavior
> wasn't well defined.

OK, I see the problem.  Briefly put, if the symlink chain doesn't
point to a non-symlink then there is no such thing as the canonical
path of the non-symlink.

Take pathname a/b/c/.../y/z.  To convert this into the canonical
path we step through the components one by one, removing "."
and "*/..".  When we reach a symlink we replace it with its value
and check again.  If at some step we arrive at a nonexistent then
we fail because the pathname is ambiguous: someone could put a
directory there or a symlink to some other location; these would
result in different canonical paths of the final non-symlink.

Thus the behavior of "readlink -f" is reasonable.  It is probably
best not to change it.

What I need is something that works like "readlink -f" and
effectively disambiguates according the rule that any missing
component is replaced by a directory.  This would work exactly
as above except that instead of failing it would succeed and
return the path as it had been converted up to that point.
This is equivalent to disambiguating by assuming that the
nonexistent components would be filled in with directories.

I reckon that it would be fairly easy to add a new option to
readlink (say, -r | --recursive) that would make it behave this way.

E.g., Suppose there is no /tmp/bar but we have /tmp/foo -> bar .    
Then "readlink -f /tmp/foo" and "readlink -f /tmp/foo/xyzzy" will fail
but "readlink -r /tmp/foo" would succeed with "/tmp/bar" and 
"readlink -r /tmp/foo/xyzzy" would succeed with "/tmp/bar/xyzzy".

Does this make sense?

-- 
Thomas Hood <address@hidden>