[omniORB] maybe an omniORB designer knows

David.Chung@USPTO.GOV David.Chung@USPTO.GOV
Thu, 24 Jun 1999 10:35:10 -0400


	I appreciate your comments.  

	Unfortunately for me, I think you are right on all accounts.  

	I suppose I should settle on "manual method" as you called it.


> -----Original Message-----
> From:	c1040@azfms.com [SMTP:c1040@azfms.com]
> Sent:	Thursday, June 24, 1999 10:05 AM
> To:	Chung, David
> Cc:	omniorb-list@uk.research.att.com
> Subject:	Re: [omniORB] maybe an omniORB designer knows
> 
> 
> > 
> > 	Specifically, I want to do the following PRIOR to the entry of
> > main(), 
> > 
> > (1) create a number of global static objects and 
> > (2) then force the execution of a function that will operate on the
> static
> > objects.  .
> 
> > 	When my program starts, the constructor for each object will check
> > if it has
> > been inserted into a global linked list..  After ALL of my objects have
> been
> > created
> > and inserted into the global linked list, the function should then be
> called
> > to 
> > operate on the objects.
> 
> How do you propose to know that ALL of your objects have been created?
> Do you 'know ahead of time' (design time)?  What about maintainability of
> this? (i.e. when you add more objects, how are you planning to be sure
> your "ALL" above is still ALL)
> 
> Or perhaps that's part of your problem - how do you know when the number
> of things inserted equals 'ALL of them'.
> 
> I read a trick somewhere (I think c/c++ journal) where you, during
> compile time, create a thing which counts the number of objects (I think)
> that you've created, and that number somehow becomes a static value
> (in a way like 'static int foo=99').  Then you could use that number
> to know when you've gotten to 'ALL'. 
> 
> Does anyone remember that article?  Am I dreaming it?  ;-)
> And, perhaps more importantly - is that what you were looking for?
> 
> > 	
> > 	One solution I was thinking about involves creating and running a
> > separate thread before the entry of main().
> > Is there a way to create a separate thread and have it run prior to the
> > entry of main()?
> 
> What does this buy you?  You still need to know when the 'last'
> constructor is being called so it can call that function you want
> called.  
> 
> I suppose its too much to either over load main (does that
> even make sense?) or otherwise make 'main' call your
> final function?   (Yes, I realize this is probably a library or
> something like that where you want as much as possible to be
> automatic - but some things just work better manually ;-)/2
> 
> rc
> 
> -----
> rusty@azfms.com		rustyc@inficad.com	rusty@descomp.com
> 
> (If this message is short, then its not meant to be curt,
> its meant to be quick.  If its long, well, sorry about
> that ;-)