allow raw (ASCII) matching via REG_RAW flag, so we no longer need a compile-time flag

You no longer need to build two libraries just for supporting raw/ASCII and UTF-8
patterns.
This makes using the library a lot easier and reduces the total binary size.
Since strings are always `unsigned char *` now internally and the raw vs. UTF-8
decision is important only in a few select places, it would make no sense to
use meta-programming techniques.

The test suite has been extended and is now fixed for cases with embedded
non-printable characters.

1 files changed, 17 insertions, 8 deletions

diff --git a/regcomp.c b/regcomp.c
index 25b1a85..f082a8b 100644
--- a/regcomp.c
+++ b/regcomp.c
@@ -181,6 +181,7 @@ static struct cvec *cclass(struct vars *, const chr *, const chr *, int);
 static struct cvec *allcases(struct vars *, pchr);
 static int cmp(const chr *, const chr *, size_t);
 static int casecmp(const chr *, const chr *, size_t);
+static int casecmp_raw(const chr *, const chr *, size_t);
 /* automatically gathered by fwd; do not hand-edit */
 /* =====^!^===== end forwards =====^!^===== */
 
@@ -231,11 +232,15 @@ struct vars {
 #define	NOTE(b)	(v->re->re_info |= (b))		/* note visible condition */
 #define	EMPTYARC(x, y)	newarc(v->nfa, EMPTY, 0, x, y)
 
-#ifdef REGEX_UTF8
-#define DECODECHR(buf, c) wctomb((char *)buf, c)
-#else
-#define DECODECHR(buf, c) ((buf)[0] = (c), 1)
-#endif
+static inline size_t
+decodechr(struct vars *v, chr *buf, pchr c)
+{
+    if (v->cflags & REG_RAW) {
+	buf[0] = c;
+	return 1;
+    }
+    return wctomb((char *)buf, c);
+}
 
 /* token type codes, some also used as NFA arc types */
 #define	EMPTY	'n'		/* no token present */
@@ -445,7 +450,11 @@ compile(
     g->tree = v->tree;
     v->tree = NULL;
     g->ntree = v->ntree;
-    g->compare = (v->cflags&REG_ICASE) ? casecmp : cmp;
+    if (!(v->cflags & REG_ICASE)) {
+	g->compare = cmp;
+    } else {
+	g->compare = (v->cflags & REG_RAW) ? casecmp_raw : casecmp;
+    }
     g->lacons = v->lacons;
     v->lacons = NULL;
     g->nlacons = v->nlacons;
@@ -1491,7 +1500,7 @@ brackpart(
 	    onechr(v, c, lp, rp);
 	    return;
 	}
-	buf_len = DECODECHR(buf, c);
+	buf_len = decodechr(v, buf, c);
 	startc = element(v, buf, buf+buf_len);
 	NOERR();
 	break;
@@ -1536,7 +1545,7 @@ brackpart(
 	switch (v->nexttype) {
 	case PLAIN:
 	case RANGE:
-	    buf_len = DECODECHR(buf, v->nextvalue);
+	    buf_len = decodechr(v, buf, v->nextvalue);
 	    NEXT();
 	    endc = element(v, buf, buf+buf_len);
 	    NOERR();