Re: Minor simplification in byte-opt.el

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> We currently do allow it (in the interpreter and the byte-compiler), but
>> I think this is a mistake and we should emit a warning (if not an error)
>> when the byte-compiler encounters such a construct.
>
> Yes, sorry, I meant that I wasn't sure whether could be synthesised (by
> cconv) without the user actually ever feeding an explicit 'closure'
> construct into the compiler. After all, 'closure' isn't really in the
> language, is it?

No, indeed, it's not in the language.  The way it usually shows up is
when a macro stuffs a *value* in there, e.g.

    `(,(get foo 'frobbing-function) ...)

The more correct alternative

    `(funcall ',(get foo 'frobbing-function) ...)

tends to suffer from false-positive warnings about quoted lambdas when
the `frobbing-function` was not byte-compiled and comes from a file that
still doesn't use lexical-binding.



        Stefan