bug#14422: 24.3; Eager Macro Expansion

[Stefan Monnier]

> Here's the revised test case that compiles cleanly and still has the
> same problem:


> (defvar eme-ll nil)

> (defmacro one (p)
>   `(progn (push ',p eme-ll)))

> (defmacro two (p)
>   (let (pp)
>     (setq pp (append eme-ll p))
>   `(progn (push ',pp eme-ll))))

> (provide 'eme)


> (require 'eme)
> (require 'ert)

 
> (ert-deftest surprise ()
>     (should
>      (equal '((one . two) one)
>           (progn
>             (one one)
>             (two two)
>             eme-ll))))

I see the test fails, but that's just because the test is wrong.
Try to create a new file foo.el:

   (require 'eme)
   
   (message "Result = %s"
            (progn
              (one one)
              (two two)
              eme-ll))

Then byte-compile it.  Then do

   emacs23 --batch -Q -l ~/tmp/foo.el
and
   emacs23 --batch -Q -l ~/tmp/foo.elc

You'll see that your code behaves differently when byte-compiled.


        Stefan


Analysis:

  (one one)

will add `one' to eme-ll at run-time.

  (two two)

reads the macroexpansion-time (e.g. compilation-time, load-time, or
run-time) value of eme-ll and adds it to eme-ll at run-time.

  eme-ll

returns the run-time value of eme-ll.