First, what you're proposing would be acceptable to me, but seems
excessive. Why is it that this is the only place in JBoss where there's
a variable to control logging? Why does everyplace else log through JMX
and does what the server administrator says should happen?
My stake in the ground: A system I can troubleshoot is infinately better
that a faster system that I can't troubleshoot.
"Schaefer, Andreas" wrote:
>
> For the binary distribution we have to set the FINAL flag to a production
> value. But when the user takes the source he/she can then compile jBoss
Fine, if 'a production value' is true. Users by and large aren't going
to be getting source (if we're doing our job right).
> to what they like or need. I don't know how much we gain but without the
> final flag you can never figure it out.
So run a benchmark n times, make it final, and re-run the benchmark n
times. Don't assume that a common optimization will make a difference in
every case. Measure.
>
> It also would help to add more logging than just for production and direct
> this through the final flag.
Or add a 'Trace' level.
