hoaparse: fix parsing of LBTT with non 0-based states

* src/hoaparse/hoaparse.yy: Here.
* src/tgbatest/lbttparse.test: Add test cases.
* src/tgbatest/hoaparse.test: Adjust.

src/tgbatest/hoaparse.test

@@ -84,6 +84,7 @@ State: 1 {0}
 -1
 EOF
 
+diff='different state numbers have been used'
 expecterr input <<EOF
 input:8.5: state number is larger than state count...
 input:2.1-9: ... declared here.
@@ -91,10 +92,7 @@ input:9.8: state number is larger than state count...
 input:2.1-9: ... declared here.
 input:10.5: state number is larger than state count...
 input:2.1-9: ... declared here.
-input:13.1: state number is larger than state count...
-input:12.1: ... declared here.
-input:18.1: state number is larger than state count...
-input:12.1: ... declared here.
+input:12.1-19.2: 2 states have been declared, but 3 $diff
 EOF
 
 cat >input <<EOF

src/tgbatest/lbttparse.test

@@ -100,6 +100,48 @@ cat >input <<EOF
 -1 4 0 1 -1 4 ! p1
 -1 5 0 1 -1 5 t
 -1
+/* this one show that state numbers do not always start from 0 */
+1 1t
+1 1
+1 -1 !p0
+1 0 -1 p0
+-1
+/* This is the output of 'lbt' on the formula 'U p0 p1', but with
+   states 1 and 2 changed to 100 and 200 */
+4 1
+0 1 -1
+100 p0
+200 p1
+-1
+100 0 -1
+100 p0
+200 p1
+-1
+200 0 0 -1
+3 t
+-1
+3 0 0 -1
+3 t
+-1
+/* This is the output of 'lbt' on the formula 'U p0 p1', but with
+   states 0 and 2 changed to 100 and 200.  This make sure the renaming
+   also applies to the initial state.
+ */
+4 1
+100 1 -1
+1 p0
+200 p1
+-1
+1 0 -1
+1 p0
+200 p1
+-1
+200 0 0 -1
+3 t
+-1
+3 0 0 -1
+3 t
+-1
 EOF
 
 run 0 ../../bin/autfilt --stats '%s %t %e %a' input > output
@@ -108,6 +150,9 @@ cat >expected<<EOF
 4 16 6 1
 1 0 0 3
 6 20 9 1
+1 2 2 1
+4 16 6 1
+4 16 6 1
 EOF
 
 diff output expected