An MRE or other hints for this macro / dynamic scoping bug?

[Dave Goel <deego3@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1309 bytes --]

Hey Emacsers,

I'm likely losing my mind. Is this not a bug, below?

I tried discussing on #emacs first, and then reported (a slightly more
complex version). But, it's perhaps not obvious, and it  was suggested I
try to produce a simpler example.

I realize I should have discussed it here first, before reporting it.  A
discussion here would probably produce a much better bug report in any
case.

Do you guys see an obvious bug here? Any simpler versions you can think
of?  Any hints where the bug could lie?

Basically, I define a macro that hardcodes the value of our variable ii
that existed before the macro definition.

And, then try to use it.


(progn
  (setq lexical-binding nil)

  (dotimes (ii 10)
    (defmacro mac ()
      `(message "%S" ,ii)
      )
    (mac)))


That's the code.
Eval it once. Works.
Eval it again. Works.

Eval it the third time. Debugger entered. Tested in emacs -Q in 28 and 29.
[[a]]

In fact, before you eval it the third time, if you change ii to jj, it
still complains about ii. [[b]]

Bugs are the behaviors  (a) and (b) above.

The macro expands to (list 'message "%S" ii). That should have worked,
right? Because ii was already set at the time of definition?

Here's the debugger behavior -
Debugger entered--Lisp error: (void-variable ii)
  (list 'message "%S" ii)

[-- Attachment #2: Type: text/html, Size: 2064 bytes --]