EXTERNAL_LIST_LOOP_2 interface and semantics changed - propose reversion

EXTERNAL_LIST_LOOP_2(elt, list) used to require that the user declare elt and list. After Ben's recent changes to the LIST_LOOP macros, the macro itself declares elt. I think this is a mistake and propose reversion. This is a similar issue to the question of how long the scope of an identifier declared in a `for' initializer lasts in C++ or C99. Some users of EXTERNAL_LIST_LOOP_2 may want to use the value of elt after the loop has exited. And there are other reasons why the user may want a different scope for elt. One could provide two versions of EXTERNAL_LIST_LOOP_*, but I think it is clearly not worth the confusion. The saving of one declaration Lisp_Object elt; is not worth the loss of flexibility and increase in magicality. Ohhhhhh, I see now Ben *has* created _NO_DECLARE versions of these macros. I think this is very subtle and confusing. EXTERNAL_LIST_LOOP_2_NO_DECLARE declares *some* of the variables, but not all. The name suggests to me "No Declarations At All", and I was suprised to discover it does some of them *and* still requires the added braces. I think this is too subtle a distinction and propose reversion to the old interface, and removal of all the _NO_DECLARE macros. With my previous implementation of these macros, it was very simple for the user to know which variables had to be declared - all the variables mentioned in the macro call had to be declared. This makes the use of these macros somewhat less magical. There's still the magic of having to provide a { } block around the call, but that seems unavoidable, and the compiler error in case of a mistake means that no subtle bugs will be introduced. This issue is similar in spirit to the TO_EXTERNAL_FORMAT debate. I want to avoid a proliferation of APIs that cause confusion for the maintainers. Both of these changes have confused *me*, the original implementor - I now feel less sure when using these macros. Martin