talloc -- Eureka*

[Sam Liddicott]

*I've got it
LIKELY(&it != NULL)
(Waits /very/ briefly for laughter and groans to die down).

The original problem I was trying to solve was that of saving country 
bumpkins from having their talloc_reference's freed from behind their back.

It never occurred to me that there were city slickers who depend on that 
sort of thing; but Samba4 seems full of it.

And yet we still have innocent simple country folk like me and Volker 
who want to take a talloc_reference and keep it.

The answer is to have two kinds of talloc_reference call (for the time 
being).

Let the old talloc_reference remain, it returns references that will be 
freed as per the dreadful current behaviour OR which cause tridges new 
error if there is more than one of THIS TYPE of parent when talloc_free 
is called.

Let there be a new talloc_safe_reference which marks the reference to 
prevent talloc_free from considering it, and which is NOT counted with 
tridges new test.

Those will be the symbol names in the talloc.so



Let us use #define so we can call talloc_new_reference as 
talloc_reference() and call the old talloc_reference as 
talloc_sinful_reference() or something.

Let us search and replace all uses of talloc_reference to become 
talloc_sinful_reference.

Then we start the process of moving talloc_sinful_reference calls to 
talloc_reference just as quickly as we feel like it.

Country bumpkins (like me) who aren't into painful surprises can use the 
new talloc_reference (macro) quite safely as it maps to 
talloc_safe_reference/

The old Samba4 code can carry on it's merry way using 
talloc_sinful_reference whose symbol is talloc_reference (for 
Metze-friendly .so compatibilty).


As a final note, maybe some mem_ctx can set a flag indicating that 
references that are made to that parent should be the non-sinful (or 
sinful) kind.


I think that this way we can preserve the old samba4 semantics while 
making safety and security available now for those who need it.


Sam