Re: [patch] [bug] src/eval.c: give Fcond Scheme-style `=>' syntax.

Karl> I've given the `cond' special form the r{4,5}rs scheme-style `=>' Karl> syntax. It's working, when the test code is run interpreted, but not Karl> when run byte-compiled. Changing a subr that has its own bytecode is at least twice as hard as changing one that doesn't. Karl> Attached is the simple test code followed by my first C code patch to Karl> XEmacs. (not counting typos in docstrings or comments.) Karl> My question is: Should the byte compiler be compiling `cond' like Karl> that, since it's in C already? Wouldn't it run faster if it only Karl> translated things like the anonymous lambda, and left the rest for Karl> the C code to run? It's reasonable for it do SOMETHING, e.g. at least check for syntactic correctness: (cond ((t . t))) => Argument list must be nil-terminated: t This is something the subr has to check for that the generated bytecode doesn't, because the byte-compiler checked it at compile time. Martin