commit d06d6aeb4fa8728f00ec1c65372e083ef15bdc1e
from: Sven M. Hallberg <pesco@khjk.org>
date: Thu Mar 24 19:13:59 2022 UTC

switch to unsigned char


note: char can be portably converted to unsigned char (modulo arithmetic)
but the converse is not true (undefined behavior)!

commit - 1ffbb49cc704cab9fe4226cc3e2ea0f95cb1b4e3
commit + d06d6aeb4fa8728f00ec1c65372e083ef15bdc1e
blob - 271ffcea40c0d6daeceae6d7df3da694ff0a19d2
blob + d9f741cecdf57b87a0fe389aac6ebfa3c13c6d4b
--- nfa.c
+++ nfa.c
@@ -62,21 +62,21 @@ range_(int x, int y)				/* [x, y) */
 }
 
 NFA
-range(char min, char max)
+range(unsigned char min, unsigned char max)
 {
 	assert (min <= max);
 	return range_(min, max + 1);
 }
 
 NFA
-nrange(char min, char max)
+nrange(unsigned char min, unsigned char max)
 {
 	assert (min <= max);
 	return range_(max + 1, min);		/* negative range */
 }
 	
 NFA
-chr(char x)
+chr(unsigned char x)
 {
 	return range(x, x);
 }
@@ -84,14 +84,14 @@ chr(char x)
 NFA
 any(void)
 {
-	return range_(CHAR_MIN, OUTPUT);	/* any input symbol */
+	return range_(0, OUTPUT);		/* any input symbol */
 						/* i.e. everything up to END */
 }
 
 NFA
 symbol(int x)					/* input only */
 {
-	assert (x >= CHAR_MIN);
+	assert (x >= 0);
 	assert (x < OUTPUT);
 	return range_(x, x + 1);
 }
@@ -136,7 +136,7 @@ think(int x)
 	 * iteration in epsilon_closure() below.
 	 */
 	assert (x >= INNER);
-	assert (x < CHAR_MIN);
+	assert (x < 0);
 	return range_(x, x);
 }
 
@@ -772,10 +772,11 @@ nfastep(struct prep *pr, int x, struct set *act, struc
 int
 nfacont(struct prep *pr, const char *input, size_t sz)
 {
-	int i, r;
+	int i, r, x;
 
 	for(i = 0; i < sz; i += r) {
-		r = nfastep(pr, input[i], &pr->act, &pr->next);
+		x = (unsigned char)input[i];
+		r = nfastep(pr, x, &pr->act, &pr->next);
 		if (r == -1)			/* no match */
 			break;
 		pr->pos += r;
blob - 34a70ddea66684a948f180773f4061be2d0cf89c
blob + 90937283569bbf8a3d2a6f72b575ecde791bd44b
--- nfa.h
+++ nfa.h
@@ -2,19 +2,19 @@
  * pesco 2021-2022, isc license
  */
 
-#include <limits.h>	/* INT_MAX, INT_MIN */
+#include <limits.h>	/* UCHAR_MAX, INT_MAX, INT_MIN */
 #include <stdlib.h>	/* size_t */
 
-#if CHAR_MAX == INT_MAX	/* it could happen */
-#error need INT_MAX > CHAR_MAX
+#if UCHAR_MAX >= INT_MAX	/* it could happen */
+#error need INT_MAX > UCHAR_MAX
 #endif
 
 /*
- * Note: INT_MAX > CHAR_MAX implies sizeof(int) > sizeof(char), and since
+ * Note: INT_MAX > UCHAR_MAX implies sizeof(int) > sizeof(char), and since
  * sizeof(char) = 1, the number of bits in an int is n * CHAR_BIT with n > 1.
  * In other words, int has at least CHAR_BIT bits more than char. CHAR_BIT is
  * greater than 1 (actually, at least 8), so it follows that the number
- * INT_MAX / 2 is still greater than CHAR_MAX (by at least 7 bits).
+ * INT_MAX / 2 is still greater than UCHAR_MAX (by at least 7 bits).
  */
 #define END	(INT_MAX / 2)	/* special input symbol meaning end of input */
 #define OUTPUT	(END + 1)	/* start of output range */
@@ -38,9 +38,9 @@ typedef struct nfa NFA;
 
 /* elementary NFAs */
 NFA fail(void);
-NFA range(char, char);
-NFA nrange(char, char);
-NFA chr(char);
+NFA range(unsigned char, unsigned char);
+NFA nrange(unsigned char, unsigned char);
+NFA chr(unsigned char);
 NFA any(void);
 NFA symbol(int);		/* input only */
 NFA output(int);