On 1/27/12 3:52 AM, Darran Lofthouse wrote:
On 01/26/2012 07:30 PM, Brian Stansberry wrote:
> On 1/26/12 8:26 AM, Ondrej Zizka wrote:
>> On Thu, 2012-01-26 at 12:54 +0530, Jaikiran Pai wrote:
>>
>>> 1) The fact that we are already having almost more than 200 message ids
>>> for EJB subsystem alone makes it look like a bit too much IMO. Looking
>>> at many of those messages, they are sometimes related to invalid
>>> arguments being passed to methods and we throwing
>>> IllegalArgumentException with a specific message id. Unfortunately, this
>>> appears to be repeated many times with different ids with a different
>>> message instead of just having a common id/message (perhaps across
>>> subsystems).
>> I'd leave it fine-grained, as it will help users find a solution for the
>> specific situation/argument. my2c
>>
>
> There are a number of places where we're reusing the same message for
> the same situation in different places (e.g. simple defensive coding
> null checks.) GSS was actually considering assigning some resources to
> convert those to separate messages. I have my doubts about the wisdom of
> that, but for sure I don't think we should spend energy now combining
> existing messages.
I think the wisdom there would be if the resulting error message logged
is ambiguous so you can't identify exactly what was null resulting in
another round trip to gather additional information when we already had
a log that could have potentially contained everything needed.
Agreed; it can certainly be helpful in some cases to have a highly
targeted message. I'll post more on this when the 7.1.0.Final pressure
is off. The topic deserves broad discussion on this list.
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat