On 2009-07-15 at 11:41:11 [+0200], Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote: > > I recently noticed that our realpath() did not comply to the POSIX > standard, and fixed that. > However, the BSD version of realpath() does not fail if only the last > path component is not accessible. While it's very simple to check if > the file really exists (just an additional [l]stat()), the other check > (if at least the parent exists) is at least slightly more involved). > > The question is, should we follow BSD style, or POSIX here? In doubt I'd go with POSIX. CU, Ingo