Re: [cp-patches] Logging fixes

[Guilhem Lavaux]

On Thu, 2005-07-14 at 13:02 +0200, Mark Wielaard wrote:
> Hi,
> 
> On Wed, 2005-07-13 at 21:54 +0200, Guilhem Lavaux wrote:
> > 2005-07-13  Guilhem Lavaux  <address@hidden>
> > 
> >         * java/util/logging/Logger.java
> >         (resetLogger): Remove all handlers from the handler list.
> > 
> >         * java/util/logging/LogManager.java
> >         (reset): Call resetLogger() too.
> >         (readConfiguration): Call reset() and use the system class
> >     loader to resolve classes.
> 

Hi Mark,

> Better would be to have a mechanism for defining the class loader to
> load the handlers (or use the Thread context class loader to do it). But
> this is probably the best we can do given the current interface. And it
> seems compatible with other implementations.
> 

Yes. That was the problem I met. However as I've pointed out in sun's
bug database, Sun has chosen to use the system class loader. I think we
must add this in our javadoc.

> > --- java/util/logging/LogManager.java   2 Jul 2005 20:32:44 -0000       1.14
> > +++ java/util/logging/LogManager.java   13 Jul 2005 18:43:41 -0000
> > @@ -443,6 +443,7 @@
> >           {
> >             logger = (Logger) ref.get();
> >  
> > +           logger.resetLogger();
> >             if (logger == null)
> >               iter.remove();
> >             else if (logger != rootLogger)
> 
> logger can be null here since it comes from a WeakReference, so put the
> resetLogger() call in the else branch.
> 
Done. Thanks.


> > This patch fixes this PR (against kaffe):
> > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=307983
> 
> Could this be turned into a mauve test that calls readConfiguration() to
> test that everything works?
> 

Yes. I'll add it asap.

> Please commit this with the fixlet pointed out above.
> 
> Thanks,
> 
> Mark

New patch attached, change log unchanged.

Guilhem.

logging_patch
Description: Text Data