Do not allow identifiers that start with a dash.

This cleans up our previous fixes for a bug whereby Bison
discarded `.field' in `$-1.field'.  The previous fixes were less
restrictive about where a dash could appear in an identifier, but
the restrictions were hard to explain.  That bug was reported and
this final fix was originally suggested by Paul Hilfinger.  This
also fixes a remaining bug reported by Paul Eggert whereby Bison
parses `%token ID -123' as `%token ID - 123' and handles `-' as an
identifier.  Now, `-' cannot be an identifier.  Discussed in
threads beginning at
<http://lists.gnu.org/archive/html/bug-bison/2011-01/msg00000.html>,
<http://lists.gnu.org/archive/html/bug-bison/2011-01/msg00004.html>.
* NEWS (2.5): Update entry describing the dash extension to
grammar symbol names.  Also, move that entry before the named
references entry because the latter mentions the former.
* doc/bison.texinfo (Symbol): Update documentation for symbol
names.  As suggested by Paul Eggert, mention the effect of periods
and dashes on named references.
(Decl Summary): Update documentation for unquoted %define values,
which, as a side effect, can no longer start with dashes either.
* src/scan-code.l (id): Implement.
* src/scan-gram.l (id): Implement.
* tests/actions.at (Exotic Dollars): Extend test group to exercise
bug reported by Paul Hilfinger.
* tests/input.at (Symbols): Update test group, and extend to
exercise bug reported by Paul Eggert.
* tests/named-refs.at (Stray symbols in brackets): Update test
group.
($ or @ followed by . or -): Likewise.
* tests/regression.at (Invalid inputs): Likewise.

9 files changed, 123 insertions, 35 deletions

diff --git a/ChangeLog b/ChangeLog
index 96f75a17..32496351 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,36 @@
+2011-01-29  Joel E. Denny  <joeldenny@joeldenny.org>
+
+	Do not allow identifiers that start with a dash.
+	This cleans up our previous fixes for a bug whereby Bison
+	discarded `.field' in `$-1.field'.  The previous fixes were less
+	restrictive about where a dash could appear in an identifier, but
+	the restrictions were hard to explain.  That bug was reported and
+	this final fix was originally suggested by Paul Hilfinger.  This
+	also fixes a remaining bug reported by Paul Eggert whereby Bison
+	parses `%token ID -123' as `%token ID - 123' and handles `-' as an
+	identifier.  Now, `-' cannot be an identifier.  Discussed in
+	threads beginning at
+	<http://lists.gnu.org/archive/html/bug-bison/2011-01/msg00000.html>,
+	<http://lists.gnu.org/archive/html/bug-bison/2011-01/msg00004.html>.
+	* NEWS (2.5): Update entry describing the dash extension to
+	grammar symbol names.  Also, move that entry before the named
+	references entry because the latter mentions the former.
+	* doc/bison.texinfo (Symbol): Update documentation for symbol
+	names.  As suggested by Paul Eggert, mention the effect of periods
+	and dashes on named references.
+	(Decl Summary): Update documentation for unquoted %define values,
+	which, as a side effect, can no longer start with dashes either.
+	* src/scan-code.l (id): Implement.
+	* src/scan-gram.l (id): Implement.
+	* tests/actions.at (Exotic Dollars): Extend test group to exercise
+	bug reported by Paul Hilfinger.
+	* tests/input.at (Symbols): Update test group, and extend to
+	exercise bug reported by Paul Eggert.
+	* tests/named-refs.at (Stray symbols in brackets): Update test
+	group.
+	($ or @ followed by . or -): Likewise.
+	* tests/regression.at (Invalid inputs): Likewise.
+
 2011-01-24  Joel E. Denny  <joeldenny@joeldenny.org>
 
 	* data/yacc.c: Fix last apostrophe warning from xgettext.
diff --git a/NEWS b/NEWS
index 593807c7..06104491 100644
--- a/NEWS
+++ b/NEWS
@@ -62,6 +62,14 @@ Bison News
 
 * Changes in version 2.5 (????-??-??):
 
+** Grammar symbol names can now contain non-initial dashes:
+
+  Consistently with directives (such as %error-verbose) and with
+  %define variables (e.g. push-pull), grammar symbol names may contain
+  dashes in any position except the beginning.  This is a GNU
+  extension over POSIX Yacc.  Thus, use of this extension is reported
+  by -Wyacc and rejected in Yacc mode (--yacc).
+
 ** Named references:
 
   Historically, Yacc and Bison have supported positional references
@@ -157,14 +165,6 @@ Bison News
   LAC is an experimental feature.  More user feedback will help to
   stabilize it.
 
-** Grammar symbol names can now contain dashes:
-
-  Consistently with directives (such as %error-verbose) and variables
-  (e.g. push-pull), grammar symbol names may include dashes in any
-  position, similarly to periods and underscores.  This is GNU
-  extension over POSIX Yacc whose use is reported by -Wyacc, and
-  rejected in Yacc mode (--yacc).
-
 ** %define improvements:
 
 *** Can now be invoked via the command line:
diff --git a/doc/bison.texinfo b/doc/bison.texinfo
index cc1e0645..8b96ad93 100644
--- a/doc/bison.texinfo
+++ b/doc/bison.texinfo
@@ -3123,12 +3123,13 @@ A @dfn{nonterminal symbol} stands for a class of syntactically
 equivalent groupings.  The symbol name is used in writing grammar rules.
 By convention, it should be all lower case.
 
-Symbol names can contain letters, underscores, periods, dashes, and (not
-at the beginning) digits.  Dashes in symbol names are a GNU
-extension, incompatible with POSIX Yacc.  Terminal symbols
-that contain periods or dashes make little sense: since they are not
-valid symbols (in most programming languages) they are not exported as
-token names.
+Symbol names can contain letters, underscores, periods, and non-initial
+digits and dashes.  Dashes in symbol names are a GNU extension, incompatible
+with POSIX Yacc.  Periods and dashes make symbol names less convenient to
+use with named references, which require brackets around such names
+(@pxref{Named References}).  Terminal symbols that contain periods or dashes
+make little sense: since they are not valid symbols (in most programming
+languages) they are not exported as token names.
 
 There are three ways of writing terminal symbols in the grammar:
 
@@ -5039,9 +5040,8 @@ Define a variable to adjust Bison's behavior.
 It is an error if a @var{variable} is defined by @code{%define} multiple
 times, but see @ref{Bison Options,,-D @var{name}[=@var{value}]}.
 
-@var{value} must be placed in quotation marks if it contains any
-character other than a letter, underscore, period, dash, or non-initial
-digit.
+@var{value} must be placed in quotation marks if it contains any character
+other than a letter, underscore, period, or non-initial dash or digit.
 
 Omitting @code{"@var{value}"} entirely is always equivalent to specifying
 @code{""}.
diff --git a/src/scan-code.l b/src/scan-code.l
index 3dd10443..66757199 100644
--- a/src/scan-code.l
+++ b/src/scan-code.l
@@ -86,7 +86,7 @@ splice	 (\\[ \f\t\v]*\n)*
    named symbol references. Shall be kept synchronized with
    scan-gram.l "letter" and "id". */
 letter	  [.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_]
-id	  -*(-|{letter}({letter}|[-0-9])*)
+id	  {letter}({letter}|[-0-9])*
 ref      -?[0-9]+|{id}|"["{id}"]"|"$"
 
 %%
diff --git a/src/scan-gram.l b/src/scan-gram.l
index 41291812..83d76506 100644
--- a/src/scan-gram.l
+++ b/src/scan-gram.l
@@ -119,7 +119,7 @@ static void unexpected_newline (boundary, char const *);
 %x SC_BRACKETED_ID SC_RETURN_BRACKETED_ID
 
 letter	  [.abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_]
-id	  -*(-|{letter}({letter}|[-0-9])*)
+id	  {letter}({letter}|[-0-9])*
 directive %{id}
 int	  [0-9]+
 
diff --git a/tests/actions.at b/tests/actions.at
index 6f267af7..24c6ac8a 100644
--- a/tests/actions.at
+++ b/tests/actions.at
@@ -158,6 +158,52 @@ AT_PARSER_CHECK([./input], 0,
 [[15
 ]])
 
+# Make sure that fields after $n or $-n are parsed correctly.  At one
+# point while implementing dashes in symbol names, we were dropping
+# fields after $-n.
+AT_DATA_GRAMMAR([[input.y]],
+[[
+%{
+# include <stdio.h>
+  static int yylex (void);
+  static void yyerror (char const *msg);
+  typedef struct { int val; } stype;
+# define YYSTYPE stype
+%}
+
+%%
+start: one two { $$.val = $1.val + $2.val; } sum ;
+one: { $$.val = 1; } ;
+two: { $$.val = 2; } ;
+sum: { printf ("%d\n", $0.val + $-1.val + $-2.val); } ;
+
+%%
+
+static int
+yylex (void)
+{
+  return 0;
+}
+
+static void
+yyerror (char const *msg)
+{
+  fprintf (stderr, "%s\n", msg);
+}
+
+int
+main (void)
+{
+  return yyparse ();
+}
+]])
+
+AT_BISON_CHECK([[-o input.c input.y]])
+AT_COMPILE([[input]])
+AT_PARSER_CHECK([[./input]], [[0]],
+[[6
+]])
+
 AT_CLEANUP
 
 
diff --git a/tests/input.at b/tests/input.at
index f223a332..8a71ff6f 100644
--- a/tests/input.at
+++ b/tests/input.at
@@ -653,17 +653,20 @@ AT_BISON_CHECK([-o input.c input.y])
 AT_COMPILE([input.o], [-c input.c])
 
 
-# Periods and dashes are genuine letters, they can start identifiers.
-# Digits cannot.
+# Periods are genuine letters, they can start identifiers.
+# Digits and dashes cannot.
 AT_DATA_GRAMMAR([input.y],
 [[%token .GOOD
          -GOOD
          1NV4L1D
+         -123
 %%
-start: .GOOD -GOOD
+start: .GOOD GOOD
 ]])
 AT_BISON_CHECK([-o input.c input.y], [1], [],
-[[input.y:11.10-16: invalid identifier: `1NV4L1D'
+[[input.y:10.10: invalid character: `-'
+input.y:11.10-16: invalid identifier: `1NV4L1D'
+input.y:12.10: invalid character: `-'
 ]])
 
 AT_CLEANUP
diff --git a/tests/named-refs.at b/tests/named-refs.at
index 74549c6e..3c7b072d 100644
--- a/tests/named-refs.at
+++ b/tests/named-refs.at
@@ -446,13 +446,14 @@ AT_SETUP([Stray symbols in brackets])
 AT_DATA_GRAMMAR([test.y],
 [[
 %%
-start: foo[ /* aaa */ *&-+ ] bar
+start: foo[ /* aaa */ *&-.+ ] bar
   { s = $foo; }
 ]])
 AT_BISON_CHECK([-o test.c test.y], 1, [],
 [[test.y:11.23: invalid character in bracketed name: `*'
 test.y:11.24: invalid character in bracketed name: `&'
-test.y:11.26: invalid character in bracketed name: `+'
+test.y:11.25: invalid character in bracketed name: `-'
+test.y:11.27: invalid character in bracketed name: `+'
 ]])
 AT_CLEANUP
 
@@ -570,23 +571,27 @@ AT_DATA([[test.y]],
 %%
 start:
   .field { $.field; }
-| -field { @-field; }
 | 'a'    { @.field; }
-| 'a'    { $-field; }
 ;
 .field: ;
--field: ;
 ]])
 AT_BISON_CHECK([[test.y]], [[1]], [],
 [[test.y:4.12-18: invalid reference: `$.field'
 test.y:4.13:        syntax error after `$', expecting integer, letter, `_', `@<:@', or `$'
 test.y:4.3-8:       possibly meant: $[.field] at $1
-test.y:5.12-18: invalid reference: `@-field'
+test.y:5.12-18: invalid reference: `@.field'
 test.y:5.13:        syntax error after `@', expecting integer, letter, `_', `@<:@', or `$'
-test.y:5.3-8:       possibly meant: @[-field] at $1
-test.y:6.12-18: invalid reference: `@.field'
-test.y:6.13:        syntax error after `@', expecting integer, letter, `_', `@<:@', or `$'
-test.y:7.12-18: invalid reference: `$-field'
-test.y:7.13:        syntax error after `$', expecting integer, letter, `_', `@<:@', or `$'
+]])
+AT_DATA([[test.y]],
+[[
+%%
+start:
+  'a' { $-field; }
+| 'b' { @-field; }
+;
+]])
+AT_BISON_CHECK([[test.y]], [[0]], [],
+[[test.y:4.9: warning: stray `$'
+test.y:5.9: warning: stray `@'
 ]])
 AT_CLEANUP
diff --git a/tests/regression.at b/tests/regression.at
index a49d1df0..5e0ead03 100644
--- a/tests/regression.at
+++ b/tests/regression.at
@@ -392,7 +392,8 @@ input.y:3.14: invalid character: `}'
 input.y:4.1: invalid character: `%'
 input.y:4.2: invalid character: `&'
 input.y:5.1-17: invalid directive: `%a-does-not-exist'
-input.y:6.1-2: invalid directive: `%-'
+input.y:6.1: invalid character: `%'
+input.y:6.2: invalid character: `-'
 input.y:7.1-8.0: missing `%}' at end of file
 input.y:7.1-8.0: syntax error, unexpected %{...%}
 ]])