Re: [Wonder-disc] why does NSLog.err -> INFO in ERXNSLogLog4jBridge?

[Nick P. <np...@sp...>]

I've fiddled with this and have gotten writing to a log file working. 
However, there is still some output that the ERXLogger system does 
not catch.

If I turn on debugging for all groups and set the debug level to 2, 
occasionally I get messages from EOF like "***EODatabaseContext: 
access without lock!". This message is followed by a stack trace.

The first message is intercepted by the wonder logging system, but 
the stack trace is not! If WOOutputPath is set, this stack trace ends 
up in the log file specified in WOOutputPath.

Thought you might like to know. This complicates debugging.

I had a thought that Wonder could examine the value of WOOutputPath 
and create a log4j appender to write to the file specified. However, 
WO is doing things behind the scenes with the value of WOOutputPath, 
so perhaps it wouldn't work.

BTW, the "access without lock" messages appear to be a bug in WO 
since I am certain that I have locked down my EOEditingContexts 
before I use them, and the stack traces show that they are deep 
within some WO code that is processing a message from a Java Client.

At 3:12 PM +0200 5/23/03, Anjo Krank wrote:
>Am Freitag, 23.05.03 um 02:25 Uhr schrieb Nick Pilch:
>
>>BTW, it would be nice if ERXLogger somehow used the setting for 
>>WOOutputPath. Have you thought about that? Do you think it's even 
>>possible?
>
>No. It's up to you to supply one or more appenders, so you'd either 
>need to set the value in code or in one of the properties files.
>
>>I'd rather let our admin handle where to write the logs, and 
>>setting the output path in JavaMonitor is an easy and supported way 
>>for him to do it. Any pointers would be welcome.
>
>He could add a "-Dlog4j.appender.R.File=/tmp/somelog.log" to his app 
>parameters. Dunno if this works, though.
>
>Cheers, Anjo


-- 

Nick Pilch / np...@sp...