On 2008-12-07 at 16:31:43 [+0100], Adrian Panasiuk <adek336@xxxxxxxxx> wrote: > It seems the access times are pretty useless here, as they're changed > when stat or ls are called. > > stat file, ls file --time=access -l report the current time as the access > time. > > This is incorrect, right? As a reference, on Linux stat and ls don't > bump the access time. AFAIK BFS's on disk structure doesn't support an access time. Apparently Axel decided to just return time(NULL), which might not be the best choice. But I don't have any really bright idea either. Maybe one could maintain the access time correctly while the inode is in memory and just forget it when it's evicted. When the inode is loaded from disk, the access (and change) time could be initialized with the modified time. That would probably result in a more consistent behavior. CU, Ingo