From: hlu@irl.eecs.wsu.edu (HJ Lu) Subject: Re: utile-1.5 problem with syslogd Date: Fri, 21 May 1993 16:47:10 GMT
In article <93141.093116NU013809@NDSUVM1.BITNET> NU013809@NDSUVM1.BITNET (Greg Wettstein) writes:
>The world probably needs another version of syslogd about as much as MS-DOS
>needs another point-and-shoot directory lister but here goes anyway....
>
>Stephen Tweedie and I spent some time earlier this year working out the
>bugs in the Berkeley derived syslogd sources. We addressed a number of issues
>in this port including:
>
> 1.) How to handle kernel logging.
>
> We removed it from syslogd and came up with something called klogd
> which acts as a client of syslogd. This utility is smart enough
> to decide whether or not the /proc or the kernel based sys_syslog
> facility should be used. The choice can be changed on the fly
> by poking at the klogd client with signals. In our (speaking
> entirely without Stephen's consent) this cleaned up the syslogd
> sources a bunch and tended to make things more orthogonal. It
> also addresses the problem that not all printk's found in the
> kernel are terminated the same, e.g. :r:n vs :n
>
My syslogd can handle the kernel messages. One time when I do
mk*fs -c /dev/hdxxx on a RLL drive. My /usr/adm/syslog grew
very fast. I believe syslogd in utile 1.5 can do the same. But
I have no time to check it. BTW, I have modified syslog in
libc 4.4 a little bit. Now '0' is used to separate each
message.
H.J.