On Friday 23 February 2001 10:42, you wrote:
> Actually, it is pretty easy to reproduce. Go into your favourite
> browser and access some freenet pages through fproxy. It will go up a
> little bit, processes that is, but will go up quite a bit on any
> failed requests. I am up to 46 processes on one machine and 147 on
> the other.
How does the request have to fail? Requests that fail cleanly (w/ Freenet
message RequestFailed) don't seem to increase the process count over the
long term for me.
There is one failure mode that's completely reproducible. If I run with the
security filter on and request a page that causes it to trip, the display of
the warning page causes at least one thread to be leaked.
This exception shows up in the log file:
java.lang.Exception: Feb 23, 2001 1:09:03
PM:Freenet.contrib.fproxy.HttpHandlerServlet:1556347983:Error:Client object
signalled: Freenet.contrib.fproxy.filter.FilterException: Unknown mime type
image/jpeg
at Freenet.support.StandardLogger.log(StandardLogger.java(Compiled
Code))
at
Freenet.contrib.fproxy.HttpHandlerServlet.get(HttpHandlerServlet.java:375)
at
Freenet.contrib.fproxy.HttpHandlerServlet.run(HttpHandlerServlet.java:156)
at
Freenet.contrib.fproxy.HttpHandlerServlet.service(HttpHandlerServlet.java:89)
at
Freenet.client.ServletConnectionHandler.run(ServletConnectionHandler.java:96)
at java.lang.Thread.run(Thread.java:498)
F
Reloading leaks at least a thread every time.
Here's the uri I have been using to reproduce the problem.
http://localhost:8081/MSK at BonsaiKitten1//facebw.jpg
I put a counter and some printlns in Freenet.client.BInstance.run.
I turns out that run() isn't returning for the case where the thread is
leaked.
It looks like maybe fproxy is being too cavalier about abandoning in
process requests?
-- gj
--
Web page inside Freenet:
freenet:MSK at SSK@enI8YFo3gj8UVh-Au0HpKMftf6QQAgE/homepage//
_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://www.uprizer.com/mailman/listinfo/devl
