sed: attempt to learn about hex escapes (e.g. \x27)

Somewhat predictably, software often wants to use \x27/\x24 among others so
that they can decline worrying about ugly escaping, if said escaping is even
possible. Right now, this software is using these and getting the wrong
results, as we'll interpret those as x27 and x24 respectively. Some examples
of this, when an exp-run was ran, were science/octopus and misc/vifm.

Go ahead and process these at all times.  We allow either one or two digits,
and the tests account for both.  If extra digits are specified, e.g. \x2727,
then the third and fourth digits are interpreted literally as one might
expect.

PR:		229925
MFC after:	2 weeks

usr.bin/sed/compile.c

@@ -49,6 +49,7 @@ static const char sccsid[] = "@(#)compile.c	8.1 (Berkeley) 6/6/93";
 #include <fcntl.h>
 #include <limits.h>
 #include <regex.h>
+#include <stdbool.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -365,6 +366,51 @@ semicolon:	EATSPACE();
 	}
 }
 
+static int
+hex2char(const char *in, char *out, int len)
+{
+	long ord;
+	char *endptr, hexbuf[3];
+
+	hexbuf[0] = in[0];
+	hexbuf[1] = len > 1 ? in[1] : '\0';
+	hexbuf[2] = '\0';
+
+	errno = 0;
+	ord = strtol(hexbuf, &endptr, 16);
+	if (*endptr != '\0' || errno != 0)
+		return (ERANGE);
+	*out = (char)ord;
+	return (0);
+}
+
+static bool
+hexdigit(char c)
+{
+	int lc;
+
+	lc = tolower(c);
+	return isdigit(lc) || (lc >= 'a' && lc <= 'f');
+}
+
+static bool
+dohex(const char *in, char *out, int *len)
+{
+	int tmplen;
+
+	if (!hexdigit(in[0]))
+		return (false);
+	tmplen = 1;
+	if (hexdigit(in[1]))
+		++tmplen;
+	if (hex2char(in, out, tmplen) == 0) {
+		*len = tmplen;
+		return (true);
+	}
+
+	return (false);
+}
+
 /*
  * Get a delimited string.  P points to the delimiter of the string; d points
  * to a buffer area.  Newline and delimiter escapes are processed; other
@@ -377,6 +423,7 @@ semicolon:	EATSPACE();
 static char *
 compile_delimited(char *p, char *d, int is_tr)
 {
+	int hexlen;
 	char c;
 
 	c = *p++;
@@ -412,6 +459,12 @@ compile_delimited(char *p, char *d, int is_tr)
 			}
 			p += 2;
 			continue;
+		} else if (*p == '\\' && p[1] == 'x') {
+			if (dohex(&p[2], d, &hexlen)) {
+				++d;
+				p += hexlen + 2;
+				continue;
+			}
 		} else if (*p == '\\' && p[1] == '\\') {
 			if (is_tr)
 				p++;
@@ -431,7 +484,7 @@ compile_delimited(char *p, char *d, int is_tr)
 static char *
 compile_ccl(char **sp, char *t)
 {
-	int c, d;
+	int c, d, hexlen;
 	char *s = *sp;
 
 	*t++ = *s++;
@@ -459,6 +512,10 @@ compile_ccl(char **sp, char *t)
 				*t = '\t';
 				s++;
 				break;
+			case 'x':
+				if (dohex(&s[2], t, &hexlen))
+					s += hexlen + 1;
+				break;
 			}
 		}
 	}
@@ -499,7 +556,7 @@ static char *
 compile_subst(char *p, struct s_subst *s)
 {
 	static char lbuf[_POSIX2_LINE_MAX + 1];
-	int asize, size;
+	int asize, hexlen, size;
 	u_char ref;
 	char c, *text, *op, *sp;
 	int more = 1, sawesc = 0;
@@ -563,6 +620,21 @@ compile_subst(char *p, struct s_subst *s)
 					case 't':
 						*p = '\t';
 						break;
+					case 'x':
+#define	ADVANCE_N(s, n)					\
+	do {						\
+		char *adv = (s);			\
+		while (*(adv + (n) - 1) != '\0') {	\
+			*adv = *(adv + (n));		\
+			++adv;				\
+		}					\
+		*adv = '\0';				\
+	} while (0);
+						if (dohex(&p[1], p, &hexlen)) {
+							ADVANCE_N(p + 1,
+							    hexlen);
+						}
+						break;
 					}
 				}
 			} else if (*p == c) {

usr.bin/sed/tests/sed2_test.sh

@@ -88,10 +88,39 @@ escape_subst_body()
 	atf_check -o 'inline:abcx\n' sed 's/[ \r\t]//g' c
 }
 
+atf_test_case hex_subst
+hex_subst_head()
+{
+	atf_set "descr" "Verify proper conversion of hex escapes"
+}
+hex_subst_body()
+{
+	printf "test='foo'" > a
+	printf "test='27foo'" > b
+	printf "\rn" > c
+	printf "xx" > d
+
+	atf_check -o 'inline:test="foo"' sed 's/\x27/"/g' a
+	atf_check -o "inline:'test'='foo'" sed 's/test/\x27test\x27/g' a
+
+	# Make sure we take trailing digits literally.
+	atf_check -o "inline:test=\"foo'" sed 's/\x2727/"/g' b
+
+	# Single digit \x should work as well.
+	atf_check -o "inline:xn" sed 's/\xd/x/' c
+
+	# Invalid digit should cause us to ignore the sequence.  This test
+	# invokes UB, escapes of an ordinary character.  A future change will
+	# make regex(3) on longer tolerate this and we'll need to adjust what
+	# we're doing, but for now this will suffice.
+	atf_check -o "inline:" sed 's/\xx//' d
+}
+
 atf_init_test_cases()
 {
 	atf_add_test_case inplace_command_q
 	atf_add_test_case inplace_hardlink_src
 	atf_add_test_case inplace_symlink_src
 	atf_add_test_case escape_subst
+	atf_add_test_case hex_subst
 }