Re: LWW and Delivering a NTSERVICE

Jeff Caldwell wrote:

>--- Wade Humeniuk <whumeniu@telus.net> wrote:
>
>  
>
>>Has anyone successfully made a LWW deliverable that runs as a NT Service?
>>
>>I have ported Franz's ntservice.cl to LW but have run into a a few snags.
>>First I have to deliver the executable with :multiprocessing nil, if not
>>the StartServiceCtrlDispatcher fails because of stack problems.  I think
>>it is because a multiprocessing deliverable does not run the "main"
>>delivery
>>entry point on the main stack.  If I deliver with :multiprocessing nil than
>>when I try to (mp:initialize-multiprocessing) it drops my main entry 
>>function
>>into the top-level-repl (ugh).  If I forgo multiprocessing I can get most
>>everything to work but the service-control-handler fails because (I assume)
>>it is called from another service manager thread into the LW image which 
>>does
>>not have multiprocessing enabled.
>>    
>>
>
>What do you mean by the service-control-handler failing? It seems to work OK
>for me (XP Pro SP2). I get the sleep... sleep... messages from the loop in
>'real-main.  Also, I can get multiprocessing working like this:
>  
>

You encountered the problem I am talking about below.  It happens when 
the service is stopped
and the service-control-handler is invoked as a result.

>(defparameter *so* nil)
>
>(defun main ()
>  (let ((args system:*line-arguments-list*))
>    (format t "args are ~S~%" args)
>    (let ((exepath (pop args)))
>      (if (member "/install" args :test #'equalp)
>          (progn
>            (add-service exepath)
>            (return-from main)))
>      (if (member "/remove" args :test #'equalp)
>          (progn
>            (remove-service)
>            (return-from main)))
>      (setq *so* *standard-output*)
>      (ntservice:start-service 'idle-main :init 'init)
>      t)))
>
>(defun init (args) 
>  (format t "Start parameters: ~S~%" args)
>  t)
>
>(defun real-main ()
>  (loop
>    (format *so* "real-main sleeping...~%")
>    (sleep 5)))
>
>(defun idle-main ()
>  (push '("NT Service Process" nil real-main) mp:*initial-processes*)
>  (mp:initialize-multiprocessing)
>  (loop
>   (format *so* "idle-main sleeping...~%")
>   (sleep 5)))
>
>I never see the idle-main sleeping... message. 
>  
>

Yes, this is the case.  The (mp:initialize-multiprocessing) call seems 
to create a repl loop
and the subsequent loop does not work.  Your idle-main code gives me 
some ideas
to implement my actual service, thank-you, but I am going to have to 
clear up the
stop problems (and also any other service indications).

>There still is a bug. In your original version, when I stop the service,
>there is an exception. The console window goes away quickly but I do see "no
>exception handler".  In my version, with multiprocessing, the interesting
>behavior below happens when I stop the service. I have to kill the console
>window manually.
>  
>

Yes the problem happens when I stop the service, I get a slightly 
different message. Yes the problems
all seem to revolve multi-processing issues.


>args are ("c:\\lisp\\projects\\ntservice\\testapp.exe" "arg1" "arg2" "arg3")
>ARGC 1 ARGV #<Pointer to type :POINTER = #x00146B18>
>ARGS: ("MyService")
>Start parameters: NIL
>real-main sleeping...
>real-main sleeping...
>real-main sleeping...
>Failed to find stack-no 000000ac stack 0012f8c4
>Stack dump
>0012f8c0  0012f8d4 20944291 00000000 00000010
>0012f8d0  000000ac 0012f8ec 2000c423 210bb7f4
>0012f8e0  0012f900 00147a50 0012f988 0012f95c
>0012f8f0  77deb6ca 77deb6ca 77deb6ca 00000001
>0012f900  00000000 00000000 0012fc90 20015ed2
>0012f910  20015ed2 00000001 77deb479 00146b10
>0012f920  00650070 00000000 00147a50 00000ec4
>0012f930  00000000 00000000 0012f99c 00000000
>0012f940  00000000 0012f900 00350034 0012fbac
>0012f950  77df5655 77deb390 00000000 0012fbbc
>0012f960  77deb568 00000050 0012f988 00000216
>0012f970  00000000 2025c3ac 0012fc90 00000000
>0012f980  7c910895 00000050 0000002c 00000001
>0012f990  00000000 00000014 00000000 0079004d
>0012f9a0  00650053 00760072 00630069 00000065
>0012f9b0  00000000 0012f9c4 00000000 7c9105c8
>real-main sleeping...
>real-main sleeping...
>real-main sleeping...
>real-main sleeping...
>
>
>  
>

This stack problem also shows up when I deliver the app with 
:multiprocessing t

Thanks for giving it a try, perhaps Lispworks can shed some more light
on the problem.

Wade