----- Original Message ----- From: "Adam Jack" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Saturday, July 13, 2002 10:36 AM Subject: RE: error with Beta 3
> I found that commons-logging seems to use the ClassLoader for the Logger as > a key into a Hashtable of factories. Unfortunately my environment/usage > returns 'null' for the class loader (which is valid per JDK 1.3 > documentation, my environment, to imply 'Boot ClassLoader') and the > Hashtable and common-logging code puke on this. > > I posted a message to the commons mailing list, but I've receive no reply. file a bugrep on nagoya.apache.org/bugzilla related to this; I dont know who maintains commons-logging to offer any more advice. > What I ended up doing was hacking the commons code to check for a null > ClassLoader, and replacing it with ClassLoader.getSystemClassLoader(). I am > insufficiently versed in this stuff to know if System = Boot, and/or if I > have future ClassCast worries to look forward to, but this seem to be > working for me for now. looks like a suitable workaround, but you need a java internals expert to be sure. Ted? Are you reading this? > > That said -- just about anything that makes commons-logging unhappy (like > not having a Log4j.properties in the 'right' place) seems to cause a > NullPointerException from commons-logging. It seems that commons-logging > does a poor job of logging anything if it's complex auto-configuration > mechanism fails, a painful flaw. Sounds like it is excessively brittle.