At 01:39 PM 2/1/2004 -0500, Christopher Faylor wrote: >On Sat, Jan 31, 2004 at 02:18:48PM -0500, Pierre A. Humblet wrote: >>Fortunately it is never used in the case of spawn: all handles are >>inherited, or the parent does the work (sockets). > >The one placed the handle is actually used is in >fhandler_socket::fixup_after_exec. I'd like Corinna's confirmation >before this patch is checked in.
Good idea. FWIW, I checked that one carefully. That's why I found the secret_event bug a while back. I also tested on Win95 with an old winsock. It looks like the handle might be used, but the tests for close on exec always block the paths where it is actually used. Pierre