Alexander Bluhm <alexander.bl...@gmx.net> writes: > On Mon, Dec 26, 2016 at 04:54:54PM -0700, Theo de Raadt wrote: >> I think programs should only block the absolutely critical things, and this >> is overreach. > > Yes, blocking SIGINT and SIGQUIT is not clever. I thought there > were races with SIGCHLD and SIGTERM where only one process would > survive. But races don't exist as parent and child communicate > over the socketpair and terminate if the other one dies. > > I have seen problems with SIGHUP during debugging, instead of > restarting syslogd died. I think this could also happen during > normal operation. > > - edit syslog.conf > - send SIGHUP > - syslogd execs itself > - newsyslog rotates logfile and sends SIGHUP > - syslogd dies as signal handlers are not installed yet > > So I suggest to block SIGHUP during startup.
ok jca@ -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE