bug#38872: 27.0.50; Keywords can be let-bound

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>>> I think that should use the same condition as set_internal, so that (let
>>> ((:k :k))) still works.
>> Thanks, I forgot about that case.
>
> We should also look at what the byte-compiler does with this code.
>
> I get the impression that the best/simplest course of action is to set
> `declared_special` to `true` for keywords.

So this would be just:

diff --git a/src/lread.c b/src/lread.c
index 8064bf4d0e..f465b451a9 100644
--- a/src/lread.c
+++ b/src/lread.c
@@ -4104,6 +4104,7 @@ intern_sym (Lisp_Object sym, Lisp_Object obarray, 
Lisp_Object index)
     {
       make_symbol_constant (sym);
       XSYMBOL (sym)->u.s.redirect = SYMBOL_PLAINVAL;
+      XSYMBOL (sym)->u.s.declared_special = true;
       SET_SYMBOL_VAL (XSYMBOL (sym), sym);
     }
 
Which seems to work?  "make check" still works after doing this, and the
test case now fails like it should do.  Some more context below.

But I have to admit I don't understand why this fixes the test case.  :-/

intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
{
  Lisp_Object *ptr;

  XSYMBOL (sym)->u.s.interned = (EQ (obarray, initial_obarray)
                                 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
                                 : SYMBOL_INTERNED);

  if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
    {
      make_symbol_constant (sym);
      XSYMBOL (sym)->u.s.redirect = SYMBOL_PLAINVAL;
      XSYMBOL (sym)->u.s.declared_special = true;
      SET_SYMBOL_VAL (XSYMBOL (sym), sym);
    }



-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no