* src/gram.h, src/gram.c (error_token_number): Remove, use

errtoken->number.
* src/reader.c (reader): Don't specify the user token number (2)
for $undefined, as it uselessly prevents using it.
* src/gram.h (token_number_t): Move to...
* src/symtab.h: here.
(state_t.number): Is a token_number_t.
* src/print.c, src/reader.c: Use undeftoken->number instead of
hard coded 2.
(Even though this 2 is not the same as above: the number of the
undeftoken remains being 2, it is its user token number which
might not be 2).
* src/output.c (prepare_tokens): Rename the `maxtok' muscle with
`user_token_number_max'.
Output `undef_token_number'.
* data/bison.simple, data/bison.c++: Use them.
Be sure to map invalid yylex return values to
`undef_token_number'.  This saves us from gratuitous SEGV.
* tests/conflicts.at (Solved SR Conflicts)
(Unresolved SR Conflicts): Adjust.
* tests/regression.at (Web2c Actions): Adjust.

NEWS

@@ -3,6 +3,14 @@ Bison News
 
 Changes in version 1.49a:
 
+* Undefined token
+  The undefined token was systematically mapped to 2 which prevented
+  the use of 2 from the user.  This is no longer the case.
+
+* Undefined token
+  If yylex returned a code out of range, yyparse could die.  This is
+  no longer the case.
+
 * Large grammars
   Are now supported (large token numbers, large grammar size (= sum of
   the LHS and RHS lengths).