From: james a. <jam...@se...> - 2010-03-11 12:46:46
|
good afternoon; On 2010-03-11, at 13:21 , james anderson wrote: > good afternoon; > > [ ... ] > > given the second version, in which the order in the source file is > reversed, the definitions succeed. modulo the warning due to the as- > then unknown qualifier. this although the method combination places > the same presence constraint on those which happen to be defined > integral to the defgeneric form and that one which is defined > separately. the only difference is in the group specification. thinking more about this, it seems as if the defgeneric operation attempts to precompute effective methods as soon as at least one of all "required" method groups has been defined. i conclude that because, if the control group specification is changed to require :control methods, both source load orders succeed. for this reduced example that approach works. for the actual case, there are aspects of the effective method computation which can depend on either runtime state or a full complement of reguired methods, so the :required t will not, in itself, suffice. |