>OK, following some tests I find that PostgreSQL doesn't validate a PL/pgSQL
>function when it builds it as I thought. I tend to use plain SQL functions
>which are validated so the situation you describe above cannot occur. That
>is why I figure that the best validation is to actually build each object.
I did this test: create functionA based on functionB, drop functionB,
create functionB.
functionA is not broken. Things are much more easy than I thought !!!!!
No need to use development tables for functions !!!!!
OK, I admit there are serious **holes** in my point of view.
Later, /JMP