* Steve Greenland said: > On 18-Dec-99, 16:29 (CST), Marek Habersack <grendel@vip.net.pl> wrote: > > Of course :)) - forget about my ramblings :)). I don't know why but I > > imagined that libc doesn't know the openlog() parameters the application > > used when it reopens the connection. > > Ahh. That's why I didn't think it was a big deal for syslog() to handle > the SIGPIPE (even given the execution of an openlog()), and you did. yes :) > Actually, it sounds now like it does at least trap the SIGPIPE, and thus > avoid the client program exiting, but it still doesn't log it correctly > (sending it to the console instead of syslog). It ought to make at least > one attempt after catching the signal. I didn't look into the libc sources yet, but it might seem the code doesn't really reopen correctly the connection to /dev/log outputting the messages to /dev/console instead... marek
Attachment:
pgpvWRVm5r5WZ.pgp
Description: PGP signature