[hibernate-dev] on broken usage of logger

David M. Lloyd david.lloyd at redhat.com
Wed Nov 23 11:15:50 EST 2011

The solution is really easy guys:

log.debugf("The thing I want to log is %s", thing);

In addition, as you have noticed, it is better to not use the formatting 
variants if you aren't actually formatting any strings (it doesn't make 
a difference if the message isn't logged, but if it is there is a slight 
performance boost).

As for error handling, yes, I think we should be doing somewhat better 
at that.  Normally this is handled by the internal logging layer but 
with the variations on they types of backends and the ways they are 
invoked, we should probably put this protection in at the front end as 
well.  I would accept a JBLOGGING JIRA to this effect.

On 11/23/2011 04:49 AM, Hardy Ferentschik wrote:
> Hi,
> this revives my email from Friday
> (http://lists.jboss.org/pipermail/hibernate-dev/2011-November/007307.html)
> where I noticed this problem as well.
> Generally speaking Sanne is of course right and the right logging method
> should be used,
> but does the logging framework really have to propagate the exception
> thrown by String.format?
> I don't think it has to. I don't think that logging should cause an
> application to die.
> Couldn't the Logger just catch the exception and write out the
> unformatted message optionally
> followed by an additional warning?
> --Hardy
> On Wed, 23 Nov 2011 11:34:29 +0100, Sanne Grinovero
> <sanne at hibernate.org> wrote:
>> Hi all,
>> I'm having a chat with Hardy about how we used the Logger in some
>> cases, and it turns out there's some confusion about
>> log.debugf vs log.debug
>> log.tracef vs log.trace
>> So the "F" variant is needed when we want to format a message, in this
>> case we should make sure the number of parameters matches exactly.
>> Also when the string is generated, like in the case of:
>> https://github.com/Sanne/hibernate-core/commit/3c5f8a568de99c7a6532647ab7b08c12738762ec#L36L485
>> it's very important that we do NOT use the formatting method, but just
>> a plain "log.debug( string )", otherwise it might happen at runtime
>> that the dynamically generated string contains some characters which
>> the String formatter might want interpret as placeholders, which are
>> missing. (HHH-6817)
>> In practice, it seems like a bad idea to allow a dynamically generated
>> formatting string at all; formatting strings as defined on the Logger
>> interface are safer;
>> In case a dynamically generated message is to be generated, we should
>> all make sure to pick the correct method signature; we're wondering if
>> there could be a reasonable proposal to make the checks of the
>> annotation processor stricter and validate for this, but there are
>> some (rare) valid use cases so just throwing compilation errors would
>> be nasty.
>> Sanne
>> _______________________________________________
>> hibernate-dev mailing list
>> hibernate-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/hibernate-dev


More information about the hibernate-dev mailing list