another big-ticket item for the todo list

[Brian Behlendorf <br...@organic.com>]

A cleanup of the code to check the returns of all system calls.  Turns out
none of the logfile writing is error checked, which caused us to not notice
that we weren't writing to a logfile for about 12 hours this morning.  On
solaris 2.5 we had made some changes last night to the ulimit in the shell
script that launches the servers (Cliff can post more details), with the
result that the logfiles weren't getting written - as far as we can tell this
was an OS bug, but the fact that the writes weren't checked for errors meant
that it wasn't noticed until I did a wc on the logfile and noticed it was
pretty small today.  No error_log messages, the servers were happily 
dishing out requests, etc.  

Cliff mentioned the idea of a general logging module - I like it.  The 
idea is that each error in the apache code would get assigned a 
"seriousness" value, with say "1" being the most critical ("can not 
write to error log") and "10" being the least ("connection timed out").  
The logging module - doesn't have to be a module, but should be easy to 
configure and possibly replace - would decide based on configuration 
whether it sends those messages to a log, to syslogd, perhaps even invoke 
an external process (mail page-brian@organic.com -s "the machine's 
hosed!" :)  Just like sendmail, say.

Anyways, I'll add this on the todo list.

	Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  brian@hyperreal.com  http://www.[hyperreal,organic].com/