> Good suggestion.  Another way (I did this in a C++ routine) is to keep a
> list of all the directory inodes you process.  Every time you open a new
> one, you check it against the list.  If it matches anything in the list,
> you skip that one.  Of course this would only work for traversing a
> reasonbly sized file tree.
>
> Vaughn

Was that an ordered or unordered list?  (My instinct would be to use an 
ordered or perhaps an indexed list.)

Actually what you want is a perfect hash.  Any collision means that you've 
visited the node already.  (When I last delt with this problem I was working 
in PERL and got hashes for free.)  

For a modest set of I-nodes you might use an array.  

There should also be a way to do it very nicely with a combination of array 
and bit-mask.