bug#65620: void function edebug-after

[Michael Heerdegen]

Hello Alan,

> -(defalias 'edebug-before nil
> +;; The following versions of `edebug-before' and `edebug-after' exist
> +;; to handle the error which occurs if either of them gets called
> +;; without an enclosing `edebug-enter'.  This can happen, for example,
> +;; when a macro mistakenly has a `form' element in its edebug spec,
> +;; and it additionally, at macro-expansion time, calls `eval',
> +;; `apply', or `funcall' (etc.)

I wonder whether what you say about `apply', or `funcall' is true: What
you can "call" in the macro expander is either a symbol or a function
_form_, i.e., a quoted lambda.  Maybe this quoted lambda is instrumented
by Edebug when forcing it in the debug spec, dunno, but specifying a
lambda expression in a macro call that is then called by the macro at
expansion time as quoted lambda makes no sense, one would rather make
the macro accept a form, or eval the function form in the expansion at
run-time.

So I'm not sure whether `apply' and `funcall' are really like `eval' in
this case.

Or - if the argument to funcall is a symbol - my question is what would
happen when macro expansion calls instrumented functions the normal way
(F . ARGS).  This works correctly, right?

Michael.