Coda File System

Re: CODA 5.3.20 compilation under FreeBSD 5 (stat structure problem)

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 6 Mar 2003 16:38:32 -0500
On Thu, Mar 06, 2003 at 10:37:56AM -0500, Greg Troxel wrote:
>   How many people would get terribly annoyed if the Coda client and server
>   processes don't do their own logrotation and we simply reopen the
>   logfiles whenever we receive a SIGHUP like all normal daemon processes.
>   That way we can leave it up to an external process (like logrotate) to
>   safely rotate the logs and we can drop the use of scandir.
> 
> I'm already annoyed because they don't use syslog!  There is no good
> reason not to, other than catering to lame systems.  But in that case,
> it can be a './configure --with-logfiles' to get the old (simplified)
> behavior.

:) Problem is that everything just dumps their output to stdout and
stderr which are redirected to the log files. This includes lwp/rpc2
and rvm.

And a lot of the logged data is really not that interesting. It probably
was useful to some developer(s) while they were working on a problem.
Like, why would I care that the 'SmonDaemon timer' just expired. I don't
believe many people know what smon (and mond) are, but they used to be
part of an extensive profiling system where clients and server dumped
trace data into a centralized sql database. I don't think anyone has
used that code in the past 4-5 years.

Jan
Received on 2003-03-06 16:40:50