Re: Re: GCIC Fullscreen applications

mfnoel wrote:
> I rely mostly on testing, but could you elaborate a bit on how reentrancy is a problem when recursion is OK? Want to know what to stay away from...
> 

Well..

I think because a function or module can be recursive without being 
reentrant.

***************
char foo[10];

int bar(char x)
{
	static char	z[2];
	if(x!=0)
	{
		foo[0]=0;
		z[0]=x;
		z[1]=0;
	}
	if(strlen(foo)==(sizeof(foo)-1)) return 0;
	strcat(foo,z);
	return bar(0);
}
***************

That's recursive, but not reentrant. If the task becomes interrupted and 
another task with another context invokes bar(), then unexpected results 
will occur - because both the external symbol "foo" and the internal "z" 
are shared between multiple contexts.

Task 1 invokes bar("A").. Expects foo to be a bunch of 'A's..
Task 1 is interrupt midflight
Task 2 invokes bar("B") and the function completes
Task 1 resumes and bar() completes. foo contains a bunch of 'B's

Of course, the example is totally trivial and useless - it's just an 
example.

So, to stay away from the "problem" :

- Keep information in task owned contexts (a stack or a chained list of 
buffers and registers)
- Serialize access to shared resources (lock/unlock mechanism for example).

--Ivan