On Sun, Aug 21, 2005 at 11:18:49AM -0700, Matthew Dillon wrote: > The comment should be put back in, and clarified even more then > it currently is. e.g. I don't understand why the full negative > range is allowed when only one or two directory entries are > faked up, or why the wraparound test was removed.
I removed the comment because it is simply incorrect. If there should be a comment, it should descripe why the on-disk offset in the directory is not, but the added comment didn't do that. Arguments against using it that way is the complication of code [for a secondary filesystem] and the possible security issues coming from it. The wraparound test was not removed, but coded explicitly. Checking for two types having identical values is a bad thing, because it can surprising results when both types have the same width and different signedness. Joerg
