Re: [Wrapper-user] How to catch other system events

[Leif M. <le...@ta...>]

Richard,
    As I described in my last mail, the native code is currently not 
capable of returning
the boolean value returned by the call to WrapperListener.controlEvent 
because that
call is not currently done synchronously.  This was necessary to work 
around some
stability problems with the JVM.

    A better approach to this exercise might be for you to describe in 
detail exactly
what you are trying to do and how and why you are going about doing 
it.   From
that info, I might have some ideas as to how you could possibly do 
things differently.
If you don't want to post that much detail on-list, then sent it to me 
directly.

Cheers,
Leif

   
Richard Luong wrote:

>Leif,
>
>Thanks again for your very informative response.  The custom dll, that
>would have been the calling code would have taken the return value,
>which is usually regarded as 0 being false and any other value as true,
>and actually "cast/convert" the response and use it as if it were an
>int.  But from what I'm reading from your response, this might be a
>little too much for myself to do.  But maybe I can work around this
>constraint.
>
>This brings me to my next question.  If the native code currently
>returns a BOOL, is there anyway that the java code can return a true or
>false, that way the calling native code will forward that response on?
>
>Richard.
>
>Leif Mortenson wrote:
>  
>
>>Richard,
>>I am confused. If I understand correctly then what you are saying does not
>>appear to be supported by the Windows API.
>>
>>It would be possible for the WrapperListener.controlEvent method to return
>>an integer rather than a boolean. This would be difficult as it would break
>>things for the user base. But it is technically possible. The problem is
>>what to do with that code once it has been returned to the native JNI
>>library.
>>
>>This is complicated by what follows, but normally the controlEvent would be
>>called from within a Handler callback function registered with the Windows
>>SetConsoleCtrlHandler function. This API declares that the handler
>>function should return a BOOL, not an int. That means that the calling
>>code would interpret 0 as FALSE and anything else as TRUE. I don't see
>>how returning an integer code could possibly work?
>>
>>Now. The problem is that the above call thread is not the way things are
>>actually implemented. For some reason I was finding that various versions
>>of the JVM liked to crash if a JNI call was made into the JVM from within
>>a Ctrl Handler. To work around this, the handler will simply store its last
>>known event code. The JVM then polls that code for a value to pass to the
>>WrapperListener.controlEvent. User code should never notice a
>>difference as signals are never fired at very high rates. The problem is
>>that as things are currently implemented, it is not possible to return even
>>the BOOL return value. Rather, the WrapperManager java code takes
>>responsibility of handling things like CTRL-C events. If the user code
>>chose not to handle the event then the WrapperManager does so.
>>
>>Is this what you are asking?
>>
>>If all you want to do is be able to pass user defined events to the Wrapper
>>and thus the JVM, then the Wrapper should handle that as is. Any user
>>defined signals should be passed right into the JVM via the
>>WrapperListener.controlEvent method.
>>
>>Cheers,
>>Leif
>>
>>Richard Luong wrote:
>>
>>
>>    
>>
>>>Leif,
>>>
>>>Thanks for your response. We have implemented a custom login dll that
>>>hooks into the windows system. It will send the service a system
>>>event. The service will then "respond" back to the the custom dll with
>>>a custom code. Thereby allowing pseudo-communication between the java
>>>thread being run as a service and the custom windows login dll.
>>>
>>>Is that not possible with the current code? If not, do you think it
>>>would be a hard thing for myself to implement? You are more
>>>knowledgable in this topic than I am. But is it possible for me to
>>>edit the contolEvent method to return back an integer to the call
>>>native code? I don't want to waste time coding something that will not
>>>work. Is it possible for a java method to return an integer in JNI?
>>>Thanks for your help. This seems like a great product and I really
>>>hope I can use it.
>>>
>>>Richard.
>>>
>>>
>>>
>>>      
>>>
>>>>From: Leif Mortenson <le...@ta...>
>>>>Reply-To: wra...@li...
>>>>To: wra...@li...
>>>>Subject: Re: [Wrapper-user] How to catch other system events
>>>>Date: Sun, 29 Feb 2004 22:05:46 +0900
>>>>
>>>>Richard,
>>>>The wrapperConsoleHandler function is simply a callback registered
>>>>with the
>>>>Windows SetConsoleCtrlHandler function. The handler actually returns
>>>>a BOOL.
>>>>Even if the code allowed you to return a custom code, the Windows API
>>>>would
>>>>not know what to do with it. Am I missing something?
>>>>| |
>>>>Cheers,
>>>>Leif
>>>>
>>>>Richard Luong wrote:
>>>>
>>>>
>>>>        
>>>>
>>>>>Leif,
>>>>>
>>>>>Thanks for your response. Sorry my response is somewhat late,
>>>>>however this project's priority was shifted around for me.
>>>>>
>>>>>You are correct in that Windows doesn't throw any signal for a use
>>>>>login. What I have is a dll, that hooks onto the windows login
>>>>>service and basically sends my java thread custom signals. My thread
>>>>>will then return an integer as a response. At least that's the plan.
>>>>>
>>>>>My question is, how do I reply to a system event? I see from your
>>>>>code sample below, that your wrapperConsoleHandler returns an int.
>>>>>And that int is true. That method then passes it to the JVM's
>>>>>controlEvent method, which returns void. Is there anyway that I can
>>>>>return my own custom code? Thanks.
>>>>>
>>>>>Richard.
>>>>>
>>>>>
>>>>>          
>>>>>
>>>>>>Subject:
>>>>>>Re: [Wrapper-user] How to catch other system events
>>>>>>From:
>>>>>>Leif Mortenson <le...@ta...>
>>>>>>Date:
>>>>>>Thu, 08 Jan 2004 18:59:21 +0900
>>>>>>To:
>>>>>>wra...@li...
>>>>>>
>>>>>>To:
>>>>>>wra...@li...
>>>>>>
>>>>>>
>>>>>>Richard,
>>>>>>Sorry I never answered you did I.
>>>>>>
>>>>>>The Wrapper currently catches system events in the native code,
>>>>>>Wrapper.DLL. It is taking any signal except the CTRL_BREAK_EVENT,
>>>>>>which
>>>>>>is used to generate thread dumps.
>>>>>>What is the actual system signal that your card is throwing? Normally
>>>>>>windows does not throw any signals that I know of when a user logs
>>>>>>in. I am
>>>>>>interesting in finding out how your card works because I have been
>>>>>>looking
>>>>>>for a nice clean way of telling when a user logs on for a while.
>>>>>>The next
>>>>>>version of the Wrapper allows you to poll for this information, but
>>>>>>there is
>>>>>>no signal that I could find that is fired when a user logs on to a
>>>>>>Windows system.
>>>>>>
>>>>>>Here is the native code that parses the signals. Note that I am
>>>>>>passing any
>>>>>>unknown signals off to the JVM as well so the controlEvent method
>>>>>>of your
>>>>>>WrapperListener should get them if they are thrown.
>>>>>>
>>>>>>---
>>>>>>int wrapperConsoleHandler(int key) {
>>>>>>int event;
>>>>>>
>>>>>>/* Call the control callback in the java code */
>>>>>>switch(key) {
>>>>>>case CTRL_C_EVENT:
>>>>>>event =
>>>>>>org_tanukisoftware_wrapper_WrapperManager_WRAPPER_CTRL_C_EVENT;
>>>>>>break;
>>>>>>case CTRL_BREAK_EVENT:
>>>>>>/* This is a request to do a thread dump. Let the JVM handle this. */
>>>>>>return FALSE;
>>>>>>case CTRL_CLOSE_EVENT:
>>>>>>event =
>>>>>>org_tanukisoftware_wrapper_WrapperManager_WRAPPER_CTRL_CLOSE_EVENT;
>>>>>>break;
>>>>>>case CTRL_LOGOFF_EVENT:
>>>>>>event =
>>>>>>org_tanukisoftware_wrapper_WrapperManager_WRAPPER_CTRL_LOGOFF_EVENT;
>>>>>>break;
>>>>>>case CTRL_SHUTDOWN_EVENT:
>>>>>>event =
>>>>>>org_tanukisoftware_wrapper_WrapperManager_WRAPPER_CTRL_SHUTDOWN_EVENT;
>>>>>>break;
>>>>>>default:
>>>>>>event = key;
>>>>>>}
>>>>>>if (wrapperJNIDebugging) {
>>>>>>printf("Got Control Signal %d->%d\n", key, event);
>>>>>>flushall();
>>>>>>}
>>>>>>
>>>>>>wrapperJNIHandleSignal(event);
>>>>>>
>>>>>>if (wrapperJNIDebugging) {
>>>>>>printf("Handled signal\n");
>>>>>>flushall();
>>>>>>}
>>>>>>
>>>>>>return TRUE; /* We handled the event. */
>>>>>>}
>>>>>>---
>>>>>>
>>>>>>Please most back with as much info as you can make available even
>>>>>>if you
>>>>>>get this working.
>>>>>>
>>>>>>Cheers,
>>>>>>Leif
>>>>>>
>>>>>>Richard Luong wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>>            
>>>>>>
>>>>>>>I want to hook onto the Windows Login (Gina.dll) so that I can login
>>>>>>>through an ID card. I can do the JNI stuff to talk from Java to the C
>>>>>>>dll. But I would like the C code to be able to talk back to the Java
>>>>>>>code through throwing events to my NT Service.
>>>>>>>
>>>>>>>My question is, does the Java Service Wrapper catch all events and
>>>>>>>pass them on? If that's so, then problem solved.
>>>>>>>
>>>>>>>If not, does it only pass on specific control events such as windows
>>>>>>>logoff, shutdown and ctr+C ?
>>>>>>>
>>>>>>>If so, can I modify the code to pass on all events? I know that I
>>>>>>>will
>>>>>>>have to use Integration Method #3, which is not a problem for me.
>>>>>>>However, any pointers to the code that is catching and possibly
>>>>>>>ignoring/passing on events would be helpful.
>>>>>>>
>>>>>>>Thanks.
>>>>>>>              
>>>>>>>
>>>>>>            
>>>>>>