The computation of nullable is broken: it doesn't handle empty

RHS's properly.
* tests/torture.at (GNU AWK Grammar): New.
* tests/sets.at (Nullable): New.
* src/nullable.c (set_nullable): Instead of blindly looping over
`ritems', loop over the rules, and then over their rhs's.
Work around Autotest bugs.
* src/warshall.c (bitmatrix_print): Don't use `+--+' as table
frame, because Autotest understand lines starting with a `+' as
traces from the shell.  Then, they are not processed properly.
Admittedly an Autotest bug, but we don't have time to wait for
Autotest to catch up.
* tests/regression.at (Broken Closure): Adjust to the new table
frames.
Move to...
* tests/sets.at: here.

tests/sets.at

@@ -0,0 +1,193 @@
+# Exercising Bison Grammar Sets.                      -*- Autotest -*-
+# Copyright 2001 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2, or (at your option)
+# any later version.
+
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+# 02111-1307, USA.
+
+AT_BANNER([[Grammar Sets (Firsts etc.).]])
+
+
+## ---------- ##
+## Nullable.  ##
+## ---------- ##
+
+AT_SETUP([Nullable])
+
+# At some point, nullable had been smoking grass, and managed to say:
+#
+# Entering set_nullable
+# NULLABLE
+#         'e': yes
+#         (null): no
+# ...
+
+AT_DATA([[input.y]],
+[[%%
+e: 'e' | /* Nothing */;
+]])
+
+AT_CHECK([[bison --trace input.y]], [], [],
+[[RITEM
+  'e'  (rule 1)
+  (rule 2)
+
+
+DERIVES
+	e derives
+		1: 'e' (rule 1)
+		2: (rule 2)
+
+
+Entering set_nullable
+NULLABLE
+	e: yes
+
+
+TC: Input BEGIN
+    @&t@
+   0
+  .-.
+ 0| |
+  `-'
+TC: Input END
+
+TC: Output BEGIN
+    @&t@
+   0
+  .-.
+ 0| |
+  `-'
+TC: Output END
+
+FIRSTS
+	e firsts
+		4 (e)
+
+
+FDERIVES
+	e derives
+		1: 'e'
+		2:
+
+
+Processing state 0 (reached by $)
+Closure: input
+
+
+Closure: output
+   0: . 'e'  (rule 1)
+   2: .  (rule 2)
+
+
+Entering new_itemsets, state = 0
+Entering append_states, state = 0
+Entering get_state, state = 0, symbol = 3 ('e')
+Entering new_state, state = 0, symbol = 3 ('e')
+Exiting get_state => 1
+Processing state 1 (reached by 'e')
+Closure: input
+   1: .  (rule 1)
+
+
+Closure: output
+   1: .  (rule 1)
+
+
+Entering new_itemsets, state = 1
+Entering append_states, state = 1
+transpose: input
+  0: @&t@
+
+transpose: output
+  0: @&t@
+
+]])
+
+AT_CLEANUP
+
+
+## ---------------- ##
+## Broken Closure.  ##
+## ---------------- ##
+
+# TC was once broken during a massive `simplification' of the code.
+# It resulted in bison dumping core on the following grammar (the
+# computation of FIRSTS uses TC).  It managed to produce a pretty
+# exotic closure:
+#
+# TC: Input
+#
+#    01234567
+#   +--------+
+#  0| 1      |
+#  1|  1     |
+#  2|   1    |
+#  3|    1   |
+#  4|     1  |
+#  5|      1 |
+#  6|       1|
+#  7|        |
+#   +--------+
+#
+# TC: Output
+#
+#    01234567
+#   +--------+
+#  0| 1      |
+#  1| 111    |
+#  2| 111    |
+#  3| 1111   |
+#  4| 111 1  |
+#  5| 111  1 |
+#  6| 111   1|
+#  7| 111    |
+#   +--------+
+#
+# instead of that below.
+
+AT_SETUP([Broken Closure])
+
+AT_DATA([input.y],
+[[%%
+a: b
+b: c
+c: d
+d: e
+e: f
+f: g
+g: h
+h: 'h'
+]])
+
+AT_CHECK([bison --trace input.y 2>&1 |
+          sed -n '/^TC: Output BEGIN/,/^TC: Output END/p'],
+         [0],
+[[TC: Output BEGIN
+           @&t@
+   01234567
+  .--------.
+ 0| 1111111|
+ 1|  111111|
+ 2|   11111|
+ 3|    1111|
+ 4|     111|
+ 5|      11|
+ 6|       1|
+ 7|        |
+  `--------'
+TC: Output END
+]])
+
+AT_CLEANUP