Hi, I have found the Solaris bug I mentioned yesterday, so I released a new version of syslog-ng, this time 1.1.23. The "bug" was in libol, so I released 0.1.19 of libol as well. Here's the changelog of libol: 1999-06-09 Balazs Scheidler <bazsi@balabit.hu> * wait at most 60 seconds in poll (so that pending closes are really closed in at most 60 seconds) * changed fatal() to werror() if the poll fails syslog-ng has not changed since yesterday. The exact bug on solaris has been fixed by the 2nd change above. poll failed in libol since the config file I tested syslog-ng with was referring to /dev/tty as a destination. When syslog-ng forked and became a session leader trying to poll /dev/tty failed. poll() returned an ENXIO error (which by the way is undocumented) and fatal() aborted the program. -- Bazsi PGP key: http://www.balabit.hu/pgpkey.txt, or finger bazsi@balabit.hu