Re: Regression 2.0 related to symbol aliases

[Paul Eggert]

Thanks for the bug report and test case.  I installed the following
somewhat-different patch, which I hope does the trick as well:

2005-04-12  Paul Eggert  <address@hidden>

        * src/symtab.c (symbol_make_alias): Call symbol_type_set,
        reverting the 2004-10-11 change to this function.
        (symbol_check_alias_consistency): Don't call symbol_type_set
        if the type name is already correct.
        * tests/input.at (Typed symbol aliases): New test, from Tim Van Holder.

--- src/symtab.c        16 Dec 2004 00:08:21 -0000      1.59
+++ src/symtab.c        12 Apr 2005 22:38:20 -0000      1.60
@@ -264,6 +264,7 @@ symbol_make_alias (symbol *sym, symbol *
        abort ();
       sym->number = symval->number =
        (symval->number < sym->number) ? symval->number : sym->number;
+      symbol_type_set (symval, sym->type_name, loc);
     }
 }
 
@@ -283,7 +284,7 @@ symbol_check_alias_consistency (symbol *
   if (!(this->alias && this->user_token_number == USER_NUMBER_ALIAS))
     return;
 
-  if (orig->type_name || alias->type_name)
+  if (orig->type_name != alias->type_name)
     {
       if (orig->type_name)
        symbol_type_set (alias, orig->type_name, orig->type_location);
--- tests/input.at      7 Feb 2005 23:14:55 -0000       1.31
+++ tests/input.at      12 Apr 2005 22:35:53 -0000      1.32
@@ -272,3 +272,28 @@ AT_PARSER_CHECK([./input], 0,
 ]])
 
 AT_CLEANUP
+
+
+## ---------------------- ##
+## Typed symbol aliases.  ##
+## ---------------------- ##
+
+AT_SETUP([Typed symbol aliases])
+
+# Bison 2.0 broke typed symbol aliases - ensure they work.
+
+AT_DATA_GRAMMAR([input.y],
+[[%union
+{
+  int val;
+};
+%token <val> MY_TOKEN "MY TOKEN"
+%type <val> exp
+%%
+exp: "MY TOKEN";
+%%
+]])
+
+AT_CHECK([bison -o input.c input.y])
+
+AT_CLEANUP