Re: socklog-unix problem, on netbsd

On Sat, Feb 04, 2006 at 02:40:36PM -0500, Charlie Brady wrote:
>
>On Sat, 21 Jan 2006, George Georgalis wrote:
>
>>># ktruss logger hello
>...
>>> 5343 logger   socket(0x1, 0x2, 0)                = 1
>>> 5343 logger   fcntl(0x1, 0x2, 0x1)               = 0
>>> 5343 logger   connect(0x1, 0xbdbe9420, 0xe)      Err#2 ENOENT
>>> 5343 logger   close(0x1)                         = 0
>>> 5343 logger   sendto(0xffffffff, 0xbfbfdcc0, 0x1f, 0, 0, 0) Err#9 EBADF
>>> 5343 logger   close(0xffffffff)                  Err#9 EBADF
>>> 5343 logger   socket(0x1, 0x2, 0)                = 1
>>> 5343 logger   fcntl(0x1, 0x2, 0x1)               = 0
>>> 5343 logger   connect(0x1, 0xbdbe9420, 0xe)      Err#2 ENOENT
>>> 5343 logger   close(0x1)                         = 0
>>> 5343 logger   sendto(0xffffffff, 0xbfbfdcc0, 0x1f, 0, 0, 0) Err#9 EBADF
>>> 5343 logger   close(0xffffffff)                  Err#9 EBADF
>>> 5343 logger   exit(0)
>
>Exit 0 eh? And sendo(-1, ...)? Looks like "logger" is very careless about 
>checking return values of system calls.
>

good point. and the netbsd hier man page needs updating as well as
rc.d/syslogd (it shouldn't be making symlinks).  I've some patches
in mind but got busy. looks like I may as well write a PR for
logger too ;)

When that's done I was thinking about a patch for socklog-conf to
use the right socket... it looks like it already detects NetBSD,
just uses the /dev/log when it should use /var/run/log -- the new
socket location.

// George

-- 
George Georgalis, systems architect, administrator <IXOYE><
http://galis.org/ cell:646-331-2027 mailto:[email protected]