regex(3): Interpret many escaped ordinary characters as EESCAPE

In IEEE 1003.1-2008 [1] and earlier revisions, BRE/ERE grammar allows for
any character to be escaped, but "ORD_CHAR preceded by an unescaped
<backslash> character [gives undefined results]".

Historically, we've interpreted an escaped ordinary character as the
ordinary character itself. This becomes problematic when some extensions
give special meanings to an otherwise ordinary character
(e.g. GNU's \b, \s, \w), meaning we may have two different valid
interpretations of the same sequence.

To make this easier to deal with and given that the standard calls this
undefined, we should throw an error (EESCAPE) if we run into this scenario
to ease transition into a state where some escaped ordinaries are blessed
with a special meaning -- it will either error out or have extended
behavior, rather than have two entirely different versions of undefined
behavior that leave the consumer of regex(3) guessing as to what behavior
will be used or leaving them with false impressions.

This change bumps the symbol version of regcomp to FBSD_1.6 and provides the
old escape semantics for legacy applications, just in case one has an older
application that would immediately turn into a pumpkin because of an
extraneous escape that's embedded or otherwise critical to its operation.

This is the final piece needed before enhancing libregex with GNU extensions
and flipping the switch on bsdgrep.

[1] http://pubs.opengroup.org/onlinepubs/9699919799.2016edition/

PR:		229925 (exp-run, courtesy of antoine)
Differential Revision:	https://reviews.freebsd.org/D10510

contrib/netbsd-tests/lib/libc/regex/data/meta.in

@@ -4,7 +4,9 @@ a[bc]d		&	abd	abd
 a\*c		&	a*c	a*c
 a\\b		&	a\b	a\b
 a\\\*b		&	a\*b	a\*b
-a\bc		&	abc	abc
+# Begin FreeBSD
+a\bc		&C	EESCAPE
+# End FreeBSD
 a\		&C	EESCAPE
 a\\bc		&	a\bc	a\bc
 \{		bC	BADRPT

contrib/netbsd-tests/lib/libc/regex/data/subexp.in

@@ -12,7 +12,7 @@ a(b+)c		-	abbbc	abbbc	bbb
 a(b*)c		-	ac	ac	@c
 (a|ab)(bc([de]+)f|cde)	-	abcdef	abcdef	a,bcdef,de
 # Begin FreeBSD
-a\(b\|c\)d	b	ab|cd	ab|cd	b|c
+a\(b|c\)d	b	ab|cd	ab|cd	b|c
 # End FreeBSD
 # the regression tester only asks for 9 subexpressions
 a(b)(c)(d)(e)(f)(g)(h)(i)(j)k	-	abcdefghijk	abcdefghijk	b,c,d,e,f,g,h,i,j

lib/libc/regex/Symbol.map

@@ -3,8 +3,11 @@
  */
 
 FBSD_1.0 {
-	regcomp;
 	regerror;
 	regexec;
 	regfree;
 };
+
+FBSD_1.6 {
+	regcomp;
+};

lib/libc/regex/regcomp.c

@@ -102,11 +102,14 @@ struct parse {
 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
 	bool allowbranch;	/* can this expression branch? */
 	bool bre;		/* convenience; is this a BRE? */
+	int pflags;		/* other parsing flags -- legacy escapes? */
 	bool (*parse_expr)(struct parse *, struct branchc *);
 	void (*pre_parse)(struct parse *, struct branchc *);
 	void (*post_parse)(struct parse *, struct branchc *);
 };
 
+#define PFLAG_LEGACY_ESC	0x00000001
+
 /* ========= begin header generated by ./mkh ========= */
 #ifdef __cplusplus
 extern "C" {
@@ -132,6 +135,7 @@ static void p_b_cclass(struct parse *p, cset *cs);
 static void p_b_eclass(struct parse *p, cset *cs);
 static wint_t p_b_symbol(struct parse *p);
 static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
+static bool may_escape(struct parse *p, const wint_t ch);
 static wint_t othercase(wint_t ch);
 static void bothcases(struct parse *p, wint_t ch);
 static void ordinary(struct parse *p, wint_t ch);
@@ -199,22 +203,10 @@ static char nuls[10];		/* place to point scanner in event of error */
 /* Macro used by computejump()/computematchjump() */
 #define MIN(a,b)	((a)<(b)?(a):(b))
 
-/*
+static int				/* 0 success, otherwise REG_something */
- - regcomp - interface for parser and compilation
+regcomp_internal(regex_t * __restrict preg,
- = extern int regcomp(regex_t *, const char *, int);
- = #define	REG_BASIC	0000
- = #define	REG_EXTENDED	0001
- = #define	REG_ICASE	0002
- = #define	REG_NOSUB	0004
- = #define	REG_NEWLINE	0010
- = #define	REG_NOSPEC	0020
- = #define	REG_PEND	0040
- = #define	REG_DUMP	0200
- */
-int				/* 0 success, otherwise REG_something */
-regcomp(regex_t * __restrict preg,
 	const char * __restrict pattern,
-	int cflags)
+	int cflags, int pflags)
 {
 	struct parse pa;
 	struct re_guts *g;
@@ -273,6 +265,7 @@ regcomp(regex_t * __restrict preg,
 	p->end = p->next + len;
 	p->error = 0;
 	p->ncsalloc = 0;
+	p->pflags = pflags;
 	for (i = 0; i < NPAREN; i++) {
 		p->pbegin[i] = 0;
 		p->pend[i] = 0;
@@ -345,6 +338,43 @@ regcomp(regex_t * __restrict preg,
 	return(p->error);
 }
 
+/*
+ - regcomp - interface for parser and compilation
+ = extern int regcomp(regex_t *, const char *, int);
+ = #define	REG_BASIC	0000
+ = #define	REG_EXTENDED	0001
+ = #define	REG_ICASE	0002
+ = #define	REG_NOSUB	0004
+ = #define	REG_NEWLINE	0010
+ = #define	REG_NOSPEC	0020
+ = #define	REG_PEND	0040
+ = #define	REG_DUMP	0200
+ */
+int				/* 0 success, otherwise REG_something */
+regcomp(regex_t * __restrict preg,
+	const char * __restrict pattern,
+	int cflags)
+{
+
+	return (regcomp_internal(preg, pattern, cflags, 0));
+}
+
+#ifndef LIBREGEX
+/*
+ * Legacy interface that requires more lax escaping behavior.
+ */
+int
+freebsd12_regcomp(regex_t * __restrict preg,
+	const char * __restrict pattern,
+	int cflags, int pflags)
+{
+
+	return (regcomp_internal(preg, pattern, cflags, PFLAG_LEGACY_ESC));
+}
+
+__sym_compat(regcomp, freebsd12_regcomp, FBSD_1.0);
+#endif	/* !LIBREGEX */
+
 /*
  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op,
  - return whether we should terminate or not
@@ -435,7 +465,10 @@ p_ere_exp(struct parse *p, struct branchc *bc)
 			EMIT(OEOW, 0);
 			break;
 		default:
+			if (may_escape(p, wc))
 				ordinary(p, wc);
+			else
+				SETERROR(REG_EESCAPE);
 			break;
 		}
 		break;
@@ -797,7 +830,10 @@ p_simp_re(struct parse *p, struct branchc *bc)
 			return (false);	/* Definitely not $... */
 		p->next--;
 		wc = WGETNEXT();
+		if ((c & BACKSL) == 0 || may_escape(p, wc))
 			ordinary(p, wc);
+		else
+			SETERROR(REG_EESCAPE);
 		break;
 	}
 
@@ -1094,6 +1130,55 @@ p_b_coll_elem(struct parse *p,
 	return(0);
 }
 
+/*
+ - may_escape - determine whether 'ch' is escape-able in the current context
+ == static int may_escape(struct parse *p, const wint_t ch)
+ */
+static bool
+may_escape(struct parse *p, const wint_t ch)
+{
+
+	if ((p->pflags & PFLAG_LEGACY_ESC) != 0)
+		return (true);
+	if (isalpha(ch) || ch == '\'' || ch == '`')
+		return (false);
+	return (true);
+#ifdef NOTYET
+	/*
+	 * Build a whitelist of characters that may be escaped to produce an
+	 * ordinary in the current context. This assumes that these have not
+	 * been otherwise interpreted as a special character. Escaping an
+	 * ordinary character yields undefined results according to
+	 * IEEE 1003.1-2008. Some extensions (notably, some GNU extensions) take
+	 * advantage of this and use escaped ordinary characters to provide
+	 * special meaning, e.g. \b, \B, \w, \W, \s, \S.
+	 */
+	switch(ch) {
+	case '|':
+	case '+':
+	case '?':
+		/* The above characters may not be escaped in BREs */
+		if (!(p->g->cflags&REG_EXTENDED))
+			return (false);
+		/* Fallthrough */
+	case '(':
+	case ')':
+	case '{':
+	case '}':
+	case '.':
+	case '[':
+	case ']':
+	case '\\':
+	case '*':
+	case '^':
+	case '$':
+		return (true);
+	default:
+		return (false);
+	}
+#endif
+}
+
 /*
  - othercase - return the case counterpart of an alphabetic
  == static wint_t othercase(wint_t ch);