Re: master 9613690: Raise an error when detecting old-style backquotes.

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

> I've reverted the commit, but this should really be fixed. Apparently
> Bovine generates incorrect code.

Bovine uses the Lisp reader to `read` things like "( ,@$2 )" and "( foo
,$1 (car ,@$2) )".  AFAICT it was designed for the old-style unquotes, but
it's been tweaked to work correctly when those commas are treated as
new-style unquotes.

The patch below seems to work around this problem.  Of course, as
mentioned in another email, just always interpreting those unquotes as
new-style works just as well (and with cleaner semantics).


        Stefan


diff --git a/src/lread.c b/src/lread.c
index c073fc4ce6..75c7e3ee55 100644
--- a/src/lread.c
+++ b/src/lread.c
@@ -2667,13 +2667,17 @@ read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
 {
   int c;
   bool uninterned_symbol = false;
+  bool went_through_retry = false;
   bool multibyte;
   char stackbuf[MAX_ALLOCA];
   current_thread->stack_top = stackbuf;
 
   *pch = 0;
 
+  goto skipretry;
  retry:
+  went_through_retry = true;
+ skipretry:
 
   c = READCHAR_REPORT_MULTIBYTE (&multibyte);
   if (c < 0)
@@ -3202,6 +3206,9 @@ read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
 	   of a list.  */
 	if (new_backquote_flag
 	    || !first_in_list
+            /* If there was some separation (space, comment, ....) between the
+               `(` and the `,`, we consider this is a new-style unquote.  */
+            || went_through_retry
 	    || (next_char != ' ' && next_char != '@'))
 	  {
 	    Lisp_Object comma_type = Qnil;