On Wed, Feb 21, 2007 at 04:15:45PM -0800, Paul Eggert wrote: > "Robert Millan [ackstorm]" <[EMAIL PROTECTED]> writes: > > > On Tue, Feb 20, 2007 at 03:19:57PM -0800, Paul Eggert wrote: > >> "Robert Millan [ackstorm]" <[EMAIL PROTECTED]> writes: > >> > >> > I haven't checked POSIX though, but if this is actualy a requirement > >> > maybe a POSIX_ME_HARDER hack would make sense here? > >> > >> POSIX makes no requirement. Other implementations are all over the > >> map here, so I suppose we can do what is more convenient. > > > > Well of course I believe my proposal is more convenient, but I have yet > > to hear a counter-argument. I take it this hasn't been discussed before? > > It probably has.... > > An argument for the current behavior of breaking the link is that many > systems don't let you write files that are currently being executed.
How about attempting to write it, and if that fails then unlink ? (preferably checking the errno code for this condition) -- Robert Millan ACK STORM, S.L. - http://www.ackstorm.es/ _______________________________________________ Bug-coreutils mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-coreutils
