Re: Failure to build mew

Shouldn't the mew people be involved with this discussion? SY> This sounds good. I'd only need to add (require 'poe) to one SY> .el wouldn't I, like ./mew/mew/mew.el? SJT> No, that would depend on what each individual file requires. SJT> Just put it in all of them. (require)'s should only go into those files that ... require it. SJT> If Lisp only had functions, we would not need to worry about requires SJT> at build time. (We might want to, for error checking etc, but we SJT> wouldn't need to.) The problem is macros. A macro gets evaluated SJT> twice. The first time, its arguments are treated as literals, and a SJT> new sexp is returned. _This sexp is a compile-time constant._ SJT> So macros are expanded by the bytecompiler for efficiency. The SJT> problem is that the bytecompiler provides _no_ facility for lazy SJT> evaluation of macros About the only way it could do that is to not byte-compile the code at all. And if you wanted to always use the run-time definition of a macro, you could never expand macros at compile-time anyways. You may as well not have a byte-compiler. And the user does have that option.