Is there any reason why this is done:
<snip>
GeronimoLogging.initialize(verboseArg == null || verboseArg.equals
(ARGUMENT_VERBOSE) ? GeronimoLogging.WARN : GeronimoLogging.DEBUG);
GeronimoLogging.setConsoleLogLevel(verboseArg == null ?
GeronimoLogging.INFO : verboseArg.equals(ARGUMENT_VERBOSE) ?
GeronimoLogging.DEBUG : GeronimoLogging.TRACE);
</snip>
Instead of something like:
<snip>
GeronimoLogging level = GeronimoLogging.WARN;
if (verboseArg != null) {
if (verboseArg.equals(ARGUMENT_VERBOSE)) {
level = GeronimoLogging.DEBUG;
} else if (verboseArg.equals(ARGUMENT_MORE_VERBOSE)) {
level = GeronimoLogging.TRACE;
}
}
GeronimoLogging.initialize(level);
</snip>
Also, I've noticed that we still end up allowing INFO for console
output (see the first snippet), which means we end up adding a bunch
of levels to the server's log4j configuration to limit stuff to above
INFO, which ends up limiting what goes into the log file, which I
generally expect to have a bunch more detail.
What do we really want to see on the default console? I assume this
output is generally intended for users who don't really care about a
lot of the internal muck, and really only expect to see ERROR
(possibly) WARN messages, in addition to the startup monitory and
shutdown messages.
Assuming this is correct, then we should set the default console
appenders level to either WARN or ERROR. I see that in
GeronimoLogging, the default is ERROR, though I've seen some
references to where the default should be WARN.
So... why don't we set the default to ERROR (ie.
GeronimoLogging.initialize(GeronimoLogging.ERROR)) and then set to
INFO,DEBUG,TRACE with -v, -vv, -vvv?
This should by default make the console output very terse, only
showing log messages when something bad happens (else its just
startup/shutdown messages) and then all the wonderful gory details
can be captured in the log file, w/o needing so many limiting logger
level entries in the log4j configuration.
Comments?
--jason
