[info-mcl] process crash

[Terje Norderhaug]

On Jul 20, 2009, at 4:22 PM, Alexander Repenning wrote:
> Just wondering... has anybody found any work around the process  
> problems in RMCL. At least so far I have not been able to use  
> process-run-function in a way that is NOT causing a crash.

I have found that there are some situations in which starting a  
process on RMCL leads to a crash (G4/OSX 10.4.11) while starting a  
process works fine in other situations. Perhaps by mapping it out, we  
can find patterns and clues to resolve the issue?

As I've mentioned already, evaluating a process-run-function in the  
listener typically leads to a crash:

(process-run-function "Seldom Starts" (lambda ()()))

However, sometimes even that works fine (after RMCL has been running  
for some time?).

An eval-enqueue of a process-run-function in the listener usually  
also leads to a crash:

(eval-enqueue `(process-run-function "Usually Crashes" #'(lambda ()())))

However, sometimes eval-enqueue with a process-run-function works  
fine even shortly after starting up RMCL, after which evaluating  
process-run-function in the listener suddenly also works.

Opening a New Listener from the File menu consistently works fine for  
me. It implicitly creates a new process as listed in the Processes  
inspector. So it is possible to create new processes.

Starting a process from a dialog item action function seems also to  
consistently work fine:

(add-subviews (make-instance 'window)
   (make-dialog-item 'button-dialog-item nil nil "Start Process"
      (lambda (b)
         (declare (ignore b))
         (process-run-function "OK" (lambda ()(sleep 30))))))

However, making a dialog item action function eval-enqueue the  
process-run-function usually induces a crash when clicking the button:

(add-subviews (make-instance 'window)
   (make-dialog-item 'button-dialog-item nil nil "Start Process"
      (lambda (b)
        (declare (ignore b))
        (eval-enqueue `(process-run-function "Usually Crashes"  
#'(lambda ()()))))))

An init.lisp file in the RMCL folder with the form below loads fine  
at startup. The process can be seen in the Processes inspector:

(ccl::process-run-function "Init" (lambda ()(sleep 30)))

However, load the same file from the listener and you'll likely get a  
crash:

(load "ccl:init.lisp")

Here is a simple tcp client/server that forks a new process on each  
request. Place the following in an init.lisp in the RMCL folder:

(in-package :ccl)
(require :opentransport)
(process-run-function "Server"
   (lambda ()
     (loop
       for i from 0
       do (with-open-stream (server (open-tcp-stream nil 8888))
            (process-wait (write-to-string i) (lambda ()(listen  
server)))
            (format server "Hello ~A~%" i)
            (force-output server)
            (process-run-function "crash?" (lambda ()(sleep 30)))))))

(defun connect ()
   (with-open-stream (connection (open-tcp-stream (local-interface-ip- 
address) 8888))
     (telnet-write-line connection "Hello?")
     (telnet-read-line connection)))

(add-subviews (make-instance 'window)
   (make-dialog-item 'button-dialog-item nil nil "Connect"
      (lambda (b)
        (declare (ignore b))
        (eval-enqueue #'connect))))

Click the Connect button. Bing-dada-boom...

Using telnet to connect to the server above on port 8888 also leads  
to an RMCL crash (after typing enter/return).

-- Terje 

> On Jul 9, 2009, at 5:28 PM, Terje Norderhaug wrote:
>> The most serious code modifications I am having to make relates to
>> less liberal use of processes. Spawning a new process at the wrong
>> time may take down RMCL. Starting a new process in general works
>> fine, as long as it is not as a consequence of user activity. It may
>> have to do with whether there are any event handling going on. For
>> example, evaluating a process-run-function like below in Fred or the
>> Listener reliably causes my vanilla RMCL to "quit unexpectedly" (G4,
>> OSX 10.4.11).
>>
>> (process-run-function "NeverStarts" (lambda ()()))
>>
>> -- Terje Norderhaug
>
> Prof. Alexander Repenning
>
> University of Colorado
> Computer Science Department
> Boulder, CO 80309-430
>
> vCard: http://www.cs.colorado.edu/~ralex/AlexanderRepenning.vcf
>
>
> _______________________________________________
> info-mcl mailing list
> info-mcl at clozure.com
> http://clozure.com/mailman/listinfo/info-mcl