William A. Rowe, Jr. wrote: > > [X] -1 for any release of 0.9.3 (regressed from 0.9.2?)
It's not a regression, but I'm trying to correct the fact that normal disconnects, al la >>> QUIT <<< 221 Goodbye. result in half closed sockets. The data connections both passive and active are correctly closed, and lie in TIME_WAIT for a bit, but these control connections are sitting at; tcp 1 0 ::1:21 ::1:46180 CLOSE_WAIT tcp 0 0 ::1:46180 ::1:21 FIN_WAIT2 not looking good. But reviewing the code, I don't see the flaw, so I'll end up spending tracing time on this to work out why we hadn't correctly invoked apr_lingering_close. Bother. With this fixed; and hopefully a resolution to the issue I had raised with Guenter r.e. file overwriting, and the issue I raised about the oddness with FNM_PERIOD matching, I really think we are ready for the "this looks like 1.0.0" release. Bill
