Appreciate this a lot Mark.
I'm pretty sure my previous code had a short sleep in each loop but
thankfully, in-coming data rarely exceeds the Servlet InputBuffer size of
8192 so looping is rare.
I have been trying to get to grips with the asynchronous stuff but have not
got it going yet. The concept is very appealing. Thanks for your
contributions.
On Wed, Aug 28, 2013 at 11:00 PM, Mark Thomas <ma...@apache.org> wrote:
> On 28/08/2013 09:41, Mark Thomas wrote:
> > On 27/08/2013 03:40, Vince Stewart wrote:
> >> hi all,
> >> thought I would add some progress on this topic.
> >> I have changed my method for reading from the HttpServletRequest object
> but
> >> the same warning message is thrown for every 8192 bytes read. I no
> longer
> >> regard my code to be suspect though am happy to be corrected. The
> >> application operates completely normally except for the warning
> message. The
> >> code I am using to read the input is shown below.
> >>
> >> public void doPost(HttpServletRequest httpServletRequest.....etc
> >> .......other code..............
> >> char[] cbuf=new char[8192];
> >> int i=0;
> >> int requestLength=httpServletRequest.getContentLength();
> >> BufferedReader in=httpServletRequest.getReader();
> >> StringBuilder sb=new StringBuilder(requestLength);
> >> while(sb.length()<requestLength){
> >> if(in.ready()){
> >> i=in.read(cbuf,0,reqLen);
> >> }
> >> sb.append(cbuf,0,i);
> >> }
> >> in.close();
> >> String message=sb.toString();
> >> //.....etc
> >
> > It looks like there is a possible Tomcat bug but the above code is not
> > the way to read data from the client as it puts the thread into a tight
> > loop while it is waiting for more data to arrive.
>
> I've just cleaned up the code for Tomcat 8.0.x to fix the bug that your
> example highlighted. Thanks for such a clear problem statement that made
> it very easy to track down the root cause of the issue. The fix will be
> in 8.0.0-RC2 onwards.
>
> That said, my comments about there being a better way to do what you are
> doing stand.
>
> Mark
>
> >
> > You'd be better off dropping the call to in.ready() and doing a blocking
> > read on the socket. The elapsed time should be the same (might even be a
> > little less) and your CPU consumption during the read when the client is
> > slow sending data will be significantly lower. If you remove the call to
> > in.ready(), I'm fairly sure you'll see the warnings disappear.
> >
> > Ideally, you should look at non-blocking IO as supported by Servlet 3.1
> > but that might be too big a change as it fundamentally changes how data
> > is read and written.
> >
> > Mark
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
> > For additional commands, e-mail: users-h...@tomcat.apache.org
> >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: users-h...@tomcat.apache.org
>
>
--
Vince Stewart
