tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Remy Maucherat <r...@apache.org>
Subject Re: cvs commit: jakarta-tomcat-4.0/catalina/src/share/org/apache/catalina/core StandardContext.java
Date Thu, 29 Aug 2002 08:34:12 GMT
Glenn Nielsen wrote:
> Remy Maucherat wrote:
> 
>> Glenn Nielsen wrote:
>>
>>> Capturing of stdout/stderr has been enabled in Tomcat 4.1 for 2-3 
>>> months now.
>>
>>
>>
>> I don't think it was, as RequestBase is deprecated, and is not used 
>> anymore (except in the old connectors).
>>
>>> What I committed was a bug fix.  If you revert it, capturing 
>>> stderr/stdout code
>>> will still exist, you would just put back in place a memory leak.
>>
>>
>>
>> Ok.
>>
>>> There is another place where you will need to implement your 
>>> SwallowOutput flag,
>>> in StandardWrapper.java.  It captures output from a load on startup 
>>> servlet.
>>
>>
>>
>> I don't have many problems with capturing during servlet init. It 
>> makes more sense to me to capture that data (no rational explanation, 
>> just a feeling).
>> It's not done during shutdown, BTW.
>>
> 
> I can add that. :-)
> 
> 
>> I can also make that optional using the same flag.
>>
>>> The thread sync issue is a good point, but this can be improved by 
>>> changing
>>> the SystemLogHandler to use Thread Local variables.
>>
>>
>>
>> Thread local is quite slow, from what I saw with OptimizeIt, so I 
>> don't want to use it.
>>
> 
>  From reading the section on Thread Local Performance here:
> 
> http://www-106.ibm.com/developerworks/java/library/j-threads3.html#h15292
> 
> Thread local performance differs based on the JVM.
> 
> 1.2 -> poor
> 1.3 -> better than normal sync _if_ you have alot of thread contention.
>        In production where performance is an issue Tomcat can spawn alot
>        of threads, so Thread Local could be of benefit.
> 1.4 -> much faster than normal sync
> 
> Based on this, for the long run, I think its better to switch to Thread 
> Local.

Thanks for the information :)

> In the long run it may be better to add support for CaptureLog to the 
> base class
> for a Processor so that the thread sync issues can be avoided altogether.

So I'll add support for log capture on shutdown (I think it's as useful 
to have it as on startup). My personal preference is to leave the 
capture switched off by default for the critical path, as it would 
confuse developers (IMO).

Remy


--
To unsubscribe, e-mail:   <mailto:tomcat-dev-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-dev-help@jakarta.apache.org>


Mime
View raw message