[Bug-openmcl] Debugging crashes of dppccl

[Erik Pearson]

Hi Gary,

It looks like most of the orphaned send ports were just not being garbage 
collected unless one does an explicit call to (gc).

A simple loop which creates a thread and runs simple code, like this

(defun test-make-threads (n)
  (dotimes (i n)
    (process-run-function ""
       #'(lambda (j) (+ j j)) i)))

will expose the issue of runaway ports, if you also have top running in 
another window. In fact, you'll also see the mregs (memory regions) count 
increasing. Again, these go back down to normal levels after a manual (gc).

I'm working on a test that will show the other problems (using my two 
changes -- to semaphores and sockets -- applied to the latest cvs.)

Erik


--On Sunday, March 7, 2004 9:22 PM -0700 Gary Byers <gb at clozure.com> wrote:

>
>
> On Sun, 7 Mar 2004, Erik Pearson wrote:
>
>>
>> It looks like 2 send and one dead name port are left after each thread.
>>
>
> I'm not sure what those ports are.
>
> I absolutely have to work on something unrelated for a day or two.  If
> there's any way for you to package up a reproducible test case, that'd
> be very, very helpful.



--
Erik Pearson
Adaptations
"Adaptation: It's not just for finches anymore."