[Qitab-devel] A problem of quux-hunchentoot and its solution

[Faré]

Dear xiehao,

thank you for excellently diagnosing this issue.
I have pushed a fix that I think is more robust,
but it is wholly untested, so please tell me if it works for you.

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
My hobby of not attending meetings about recycling saves more energy than
your hobby of recycling.  — John McCarthy


On Thu, Apr 24, 2014 at 11:03 PM, 昊 谢 <xiehao at outlook.com> wrote:
> + qitab-devel
>
> On Apr 25, 2014, at 15:00 , 昊 谢 <xiehao at outlook.com> wrote:
>
> HI Francois-Rene Rideau, qitab-devel,
>
> I encountered a problem when I’m using your quux-hunchentoot and I have a
> simple way to fix it. Could you please consider it?
>
> When I’m using the thread-pooling taskmaster and starting acceptor, there
> may be an error invoked occasionally like this:
> debugger invoked on a SIMPLE-ERROR in thread
> #<THREAD "quux-hunchentoot-dispatcher-*:8080" RUNNING {10035DDF03}>:
>   The assertion
>   (EQ QUUX-HUNCHENTOOT::DISPATCHER-PROCESS
> (BORDEAUX-THREADS:CURRENT-THREAD))
>   failed with QUUX-HUNCHENTOOT::DISPATCHER-PROCESS = NIL,
>   (BORDEAUX-THREADS:CURRENT-THREAD) =
>   #<SB-THREAD:THREAD "quux-hunchentoot-dispatcher-*:8080" RUNNING
> {10035DDF03}>.
>
> I think the reason for this is quite straightforward. While the method
> “execute-acceptor” is creating dispatcher thread and set it by the
> dispatcher-process accessor of taskmaster, we can’t imagine the “setf”
> action comes first. In other words, if dispatcher thread started before the
> accessor was set, the assertion in ensure-dispatcher-process may fail.
>
> My solution for this is quite simple and here’s the effective solution,
> adding something like a spin-lock.
> diff --git a/thread-pooling.lisp b/thread-pooling.lisp
> index dcb3427..8f3f32b 100644
> --- a/thread-pooling.lisp
> +++ b/thread-pooling.lisp
> @@ -146,6 +146,7 @@ then the reply will be HTTP 503."))
>
>  (defmethod ensure-dispatcher-process ((taskmaster
> thread-pooling-taskmaster))
>    (with-taskmaster-accessors (dispatcher-process) taskmaster
> +    (do () (dispatcher-process) (sleep 0.1))
>      (assert (eq dispatcher-process (bt:current-thread)))))
>
>  (defmethod decrement-taskmaster-accept-count ((taskmaster
> thread-pooling-taskmaster))
>
> Maybe you can get a better patch to solve this problem and I’m looking
> forward your reply.
>
> Thanks & have a good day
> Hao Xie
>