tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Remy Maucherat <>
Subject Re: Proposed new CometEvent.notify method
Date Fri, 04 May 2007 13:52:00 GMT
Filip Hanik - Dev Lists wrote:
> turns out the user found a problem, what I was describing is a possible, 
> but very hard scenario to actual run into. the CoyoteAdapter->read(on a 
> client disconnect) would have to happen at the same time as 
> otherThread->Servlet->write
> but as you said, he didn't run into this scenario, he was able to 
> synchronize to avoid it, since the adapter calls close/end if there is 
> an exception upon the read.

Yes. Since the read is made outside of the user control, there is a 
major problem if it has side effects (either by the IO layer putting 
itself in error mode automagically, or more directly if the request was 
closed without telling the servlet). I briefly looked at the NIO code 
which was involved in the stack trace in JDK 6, and it doesn't seem to 
be doing funny things; the only way to run into it is if the user did 
close the channel explicitly. So I think it should work too.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message