> On Feb 1, 2026, at 07:14, Mark Johnston <[email protected]> wrote: > > On Sat, Jan 31, 2026 at 02:30:35PM -0800, Doug Hardie wrote: >> AMD machine. Just upgraded to 15.0. Everything works except for syslogd >> which crashes on startup. Debug output: >> >> sermons# /usr/sbin/syslogd -sd >> new socket fd is 6 >> shutdown >> sending on socket >> new socket fd is 7 >> shutdown >> sending on socket >> Trying peer: /var/run/log >> new socket fd is 8 >> listening on socket >> sending on socket >> Trying peer: /var/run/logpriv >> new socket fd is 10 >> listening on socket >> sending on socket >> off & running.... >> init >> cfline("*.err;kern.warning;auth.notice;mail.crit /dev/console", f, >> "-wn,wnssl,wncert", "*", "*") >> cfline("*.notice;authpriv.none;kern.debug;mail.crit;wn.crit;wnssl.crit;wncert.crit >> /var/log/messages", f, "-wn,wnssl,wncert", "*", "*") >> Failed to xfer configuration nvlist: Socket is not connected >> logmsg: pri 53, flags 0, from sermons, msg Failed to xfer configuration >> nvlist: Socket is not connected >> Logging to CONSOLE /dev/console > > The underlying problem is that "wn", "wnssl", "wncert" are not valid > syslog facilities, so the configuration parser fails. Due to a bug, it > doesn't give you a useful message in this case. Before, the parser was > less strict; we perhaps should restore the old behaviour. > > Here is a patch that partly addresses the proble: > https://reviews.freebsd.org/D55033 >
I am not convinced that passing errors like that is appropriate. Yes it used to work, but I think that was a mistake now. Syslogd should generate an error message in this situation and then possibly ignore the incorrect entries as if they were not there. -- Doug
