Re: ls -i inefficiency

[Eric Blake]

> Hmmm, won't that break the behavior on dangling symlinks?

Good catch.  I was starting to wonder about that myself.

> here's the coreutils 5.94 behavior:
> 
>    $ mkdir d
>    $ cd d
>    $ ln -s nowhere x
>    $ ls -l
>    total 0
>    lrwxrwxrwx 1 eggert eggert 7 Feb 26 01:17 x -> nowhere
>    $ ls -L
>    ls: x: No such file or directory

But the 5.2.1 behavior was different:

$ ls -l
total 0
lrwxrwxrwx  1 ericb cygwin 4 Feb 26 19:41 broken -> none
$ ls -L
broken
$ ls --version | head -n1
ls (coreutils) 5.2.1

> 
> I haven't tried the patch, but from your description of it it sounds
> like the output of the last command would be "x", which doesn't sound
> right to me.

What does Solaris 10 do?  I'll prepare a patch that adds a testsuite
case for whatever decision we make, and which repairs the regression
if we decide that the 5.94 behavior of warning about an inability to
dereference broken symlinks when -L is in affect is the correct
approach.  However, I just reread the POSIX wording, and it is silent
on the matter of broken symlinks as far as I can tell.

-- 
Eric Blake