Rossen,

I agree with your interpretation and that is how I originally implemented
it.   However there are authoritative test suites that take a different
view.....     but I'm inclined to ignore those test suites when they have
such nonsensical results, so I should have stuck with my original impl and
will now revert to the previous behaviour.

cheers




On 13 August 2013 01:37, Rossen Stoyanchev <[email protected]>wrote:

> Hi Greg,
>
> Thanks for commenting. Isn't the spec relatively clear? In other words
> that the dispatch should have no effect until the exiting thread has
> returned.
>
> I was very glad to find the qualification in the spec when I first
> came across the need for it since code in the exiting thread needs to
> be able to check reliably if async has started. If we can't rely on
> isAsyncStarted, and for example had to implement a request attribute
> solution, we'd have to also remove the attribute at some point, so
> that the subsequently dispatched thread doesn't think that async has
> started. There is no good place I can think of for doing that. It has
> to be as late as possible in the exiting thread, i.e. after all
> filters, and as early as possible in the subsequently dispatched
> thread, i.e. before all filters. Also, such a request attribute would
> be framework specific and would not mix well with filters from other
> frameworks. This is where I see the purpose of this flag as essential.
>
> Rossen
>
> On Mon, Aug 12, 2013 at 11:09 AM, Greg Wilkins <[email protected]> wrote:
> > Rossen, David,
> >
> > I recently changed the behaviour of isAsyncStarted() under protest
> because
> > it was pointed out to me that we were not following the spec.  I feared
> that
> > it would lead to exactly the kind of problem that you are having here.
> >
> > Currently I'm thinking that I will revert the default behaviour so that
> > isAsyncStarted will return true until the original dispatch is completed,
> > but add a configuration option to switch on strict adherence to the spec
> > (but I cannot think why anybody would want that).
> >
> > cheers
> >
> >
> >
> >
> >
> > On 13 August 2013 01:03, Rossen Stoyanchev <[email protected]>
> > wrote:
> >>
> >> Hi Simone,
> >>
> >> Sure, no worries. Actually David already opened a ticket for this issue:
> >> https://bugs.eclipse.org/bugs/show_bug.cgi?id=413901
> >>
> >> Maybe the title can be updated now that we have more information. Will
> >> that work?
> >>
> >> Rossen
> >> _______________________________________________
> >> jetty-users mailing list
> >> [email protected]
> >> https://dev.eclipse.org/mailman/listinfo/jetty-users
> >
> >
> >
> >
> > --
> > Greg Wilkins <[email protected]>
> > http://www.webtide.com
> > Developer advice and support from the Jetty & CometD experts.
> > Intalio, the modern way to build business applications.
> >
> > _______________________________________________
> > jetty-users mailing list
> > [email protected]
> > https://dev.eclipse.org/mailman/listinfo/jetty-users
> >
> _______________________________________________
> jetty-users mailing list
> [email protected]
> https://dev.eclipse.org/mailman/listinfo/jetty-users
>



-- 
Greg Wilkins <[email protected]>
http://www.webtide.com
Developer advice and support from the Jetty & CometD experts.
Intalio, the modern way to build business applications.
_______________________________________________
jetty-users mailing list
[email protected]
https://dev.eclipse.org/mailman/listinfo/jetty-users

Reply via email to