On Wed, Oct 22, 2008 at 11:49 PM, Emmanuel Lecharny <[EMAIL PROTECTED]> wrote: > I don't really know -yet- if it's a real problem, but the thing is > that JIRA is remanent, when a mail will be buried pretty quickly. So > having a JIRA, even if it's not a bug, is a better way to be sure that > it's addressed, even if it's in one month, as we usually check the > list of JIRAs before releasing a new version.
Fair enough. I prefer to determine if it is a real problem first, otherwise you get newbies bloating the issues tracker with non-problems. And this makes life more difficult for the maintainers. We got that problem with the Maven stuff. Trawling through JIRA for issues relating to WriteFuture I found http://issues.apache.org/jira/browse/DIRMINA-249 Which looks like it fixes my problem, or at least adds WriteFuture.setException(Throwable cause) and would avoid the deadlock I am currently getting - but its a 2.0 feature. Some of this may need back-porting. I've tried using IoFutureListener() but on exceptions operationComplete() is not fired. I've added an IoFilter to the Session chain which handles exceptionCaught() and notifies the specified WriteFuture via setWritten(false). I also had a look at SocketIo to see how it differed from DatagramIo. SocketIoProcessor.doFlush() will call scheduleRemove(session) when it catches exceptions during ch.write() - which will close the channel - but also doesn't appear to notifiy WriteFuture. I've raised a Jira https://issues.apache.org/jira/browse/DIRMINA-632 Its based on mina-2.0, i'll submit 1.0 and 1.1 patches after Monday. I've modified the example memory monitor to show the problem when you pull your network cable or disable network adapter.
