I have fixed the problem with cstructs now, but this raises an interesting question. Consider this program:
fun f: int -> int = "$1" requires A; println$ f 1; Should this compile? At first glance you'd say no, because "A" isn't available which should remove "f" and thereby break the print. But consider this: header A = ".."; header A = ".."; fun f: ... requires A; By specification, both headers are emitted if f is used. In other words the *set* of headers tagged A are emitted. In other words it isn't an error if the set contains 1 member, or two members, so why not no members? It isn't what I expected, but it is, in fact consistent! So what actually happens in the malloc example is: stdlib_h is defined in the wrong place, and so isn't found when malloc is used, but still, felix correctly emits .. nothing in support of the malloc. But the code compiles by luck, because stdlib.h is a header that is always emitted by the compiler. So that's why it works: there's no error. The only way to fix this is require exactly one unique tag. After all you can chain tags: header A = " .. "; header B = " .. "; header AB requires A, B; Not sure if I like this though. Comments? In fact, the felix requirements expressions are quite complex and general: you can use and's and or's and not's etc.. and it isn't clear what that means either :) Original I though of: proc create_thread: int = "posix::create_thread ($1);" requires POSIX; proc create_thread: int = "windows::create_thread($1);" requires WIN32; as a way of selecting bindings, with the semantics above this isn't possible. but we can always add predicates .. for example requires defined(posix); // shades of c-pre-proc ...;( This leaves in question what "and" and "or"mean.. I expected require A and B to means the same as require A,B; that is, to require both, and require A or B; to mean: try A, but if not found use B instead: a simple example of the idea is to have require epoll or poll or select; meaning use epoll if available, otherwise poll, and if that isn't available use select. however a set-wise interpretation has almost the opposite meaning: require A or B; means to require the union, which is both A and B, whereas requiring A and B would be the intersection, which would currently be empty (since one insertion can't have more than one tag). BTW: I believe the problem of resolving such dependencies with alternatives is NP-hard, which is why production package managers like Debian can't effectively allow this option. -- john skaller skal...@users.sourceforge.net ------------------------------------------------------------------------------ Increase Visibility of Your 3D Game App & Earn a Chance To Win $500! Tap into the largest installed PC base & get more eyes on your game by optimizing for Intel(R) Graphics Technology. Get started today with the Intel(R) Software Partner Program. Five $500 cash prizes are up for grabs. http://p.sf.net/sfu/intelisp-dev2dev _______________________________________________ Felix-language mailing list Felix-language@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/felix-language
