Re: GWorkspace hangs on first exit attempt. Remote objects problem?

[Riccardo Mottola]

Hi,

some further information

1) unregisterClient is (oneway void). So it is asynchronous, no need to 
wait for any value returned (Sebastian suggested doing this to me, but 
is already correct)
2) if I wrap [fswatcher unregisterClient] in an exception handler, on 
the machines which caouse problems I do get an exception.

2013-03-08 01:06:25.990 GWorkspace[7503] [GWorkspace 
shouldTerminateApplication] unregister fswatcher: <NSException: 
0x318ac38> NAME:NSInvalidReceivePortException REASON:invalidated while 
awaiting reply INFO:(null)

Doesn't it look like exit() kills fswatcher but something still needs to 
be sent? The stacktrace is:

#0  -[NSException raise] (self=0x232efb8,
    _cmd=0x7ffff6e2f9e0 <_OBJC_SELECTOR_TABLE+384>) at NSException.m:955
#1  0x00007ffff696a6f1 in +[NSException raise:format:] (
    self=0x7ffff6e2fb40 <_OBJC_Class_NSException>, _cmd=<optimized out>,
    name=0x7ffff6e63120 <_OBJC_INSTANCE_0>,
    format=0x7ffff6e1b200 <_OBJC_INSTANCE_68>) at NSException.m:835
#2  0x00007ffff69383e3 in -[NSConnection(Private) _getReplyRmc:] (
    self=0x124a7d8, _cmd=0x7ffff6e1cdc0 <_OBJC_SELECTOR_TABLE+1536>, sn=26)
    at NSConnection.m:3219
#3  0x00007ffff6933422 in -[NSConnection(GNUstepExtensions) 
forwardInvocation:forProxy:] (self=0x124a7d8, _cmd=0x7ffff6e2ad10 
<_OBJC_SELECTOR_TABLE+592>,
    inv=0x22bd328, object=0x12de0f8) at NSConnection.m:2086
#4  0x00007ffff6a7472b in GSFFIInvocationCallback (cif=0x232eee0,
    retp=0x7fffffffd7a0, args=<optimized out>, user=0x2366718)
    at GSFFIInvocation.m:629
#5  0x00007ffff1c4ad09 in ffi_closure_unix64_inner ()
   from /usr/lib64/libffi.so.6
#6  0x00007ffff1c4b054 in ffi_closure_unix64 () from /usr/lib64/libffi.so.6
#7  0x00000000004137ad in -[GWorkspace applicationShouldTerminate:] (
    self=0xd0ca48, _cmd=<optimized out>, app=<optimized out>)
    at GWorkspace.m:635
#8  0x00007ffff70daf02 in -[NSApplication terminate:] (self=0xd2be98,
    _cmd=<optimized out>, sender=<optimized out>) at NSApplication.m:3462


I see _getReplyRmc:  but there is going to be none, the app is dead and 
it is oneway void.

Does this sound plausible?

Riccardo

Riccardo Mottola wrote:
> Hi,
>
> some of you may have noticed that GWorkspace sometimes do not exit at 
> the first attempt, but at the second. It does not happen on all 
> computers, but on those where it does, it happens consistently. I and 
> Sebastian did a "NSLog debugging", my previous attempts with gdb where 
> unsuccessful.
>
> in GWorkspace, in applicationShouldTerminate, we have this code:
>
>
>   if (fswatcher) {
>     NSConnection *conn = [(NSDistantObject *)fswatcher 
> connectionForProxy];
>
>     if ([conn isValid]) {
>       [nc removeObserver: self
>                       name: NSConnectionDidDieNotification
>                     object: conn];
>       [fswatcher unregisterClient: (id <FSWClientProtocol>)self];
>       DESTROY (fswatcher);
>     }
>   }
>
> the problematic line is:
> [fswatcher unregisterClient: (id <FSWClientProtocol>)self];
>
> When this GWS hangs, it hangs because this line does not return. 
> fswatcher is the daemon object.
>
> fswatcher will exit with exit() when the last client unregisters and 
> in this case it is happening, we checked the code-path and also "ps" 
> shows that it indeed exists. Why does it hang? And why does it hang on 
> certain machines only?
>
> I fear there could be a base problem, when invoking remote objects?
>
> Riccardo
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> https://lists.gnu.org/mailman/listinfo/discuss-gnustep