On Jan 22 18:33, Pierre A. Humblet wrote:
2004-01-22 Pierre Humblet [EMAIL PROTECTED]
* fhandler_socket.cc (fhandler_socket::create_secret_event): Avoid
creating multiple handles. Always allow event inheritance but set the
handle inheritance appropriately. Improve error
On Fri, Jan 23, 2004 at 10:59:52AM +0100, Corinna Vinschen wrote:
On Jan 22 18:33, Pierre A. Humblet wrote:
2004-01-22 Pierre Humblet [EMAIL PROTECTED]
* fhandler_socket.cc (fhandler_socket::create_secret_event): Avoid
creating multiple handles. Always allow event inheritance but
Christopher Faylor wrote:
I agree, with one nit. Was there a reason for getting rid of the handle
protection in this patch? We are apparently stumbling over a problem with
handle corruption in the current CVS so removing a chance for protection
seems like we're going backwards.
The
On Fri, Jan 23, 2004 at 11:04:05AM -0500, Pierre A. Humblet wrote:
Christopher Faylor wrote:
I agree, with one nit. Was there a reason for getting rid of the handle
protection in this patch? We are apparently stumbling over a problem with
handle corruption in the current CVS so removing
fhandler_socket.cc has a handle leak: when several accept()'s are
made on an AF_LOCAL parent socket, it accumulates secret_event handles.
Also close on exec does not work properly on the secret_event when
the call occurs after the event creation.
The patch fixes both issues. For the close on