Fixed PCRE so that global matching with patterns with \b works.

3 files changed, 19 insertions, 4 deletions

diff --git a/ext/pcre/pcrelib/internal.h b/ext/pcre/pcrelib/internal.h
index 2b28ac1a54..b6680a0d50 100644
--- a/ext/pcre/pcrelib/internal.h
+++ b/ext/pcre/pcrelib/internal.h
@@ -307,6 +307,7 @@ typedef struct match_data {
   const uschar *end_subject;    /* End of the subject string */
   const uschar *end_match_ptr;  /* Subject position at end match */
   int     end_offset_top;       /* Highwater mark at end of match */
+  char   regprev;		/* Character previous to subject string (Andrey Zmievski) */
 } match_data;
 
 /* Bit definitions for entries in the pcre_ctypes table. */
diff --git a/ext/pcre/pcrelib/pcre.c b/ext/pcre/pcrelib/pcre.c
index caeeaa4971..51cb6b6c87 100644
--- a/ext/pcre/pcrelib/pcre.c
+++ b/ext/pcre/pcrelib/pcre.c
@@ -3321,8 +3321,12 @@ for (;;)
     case OP_NOT_WORD_BOUNDARY:
     case OP_WORD_BOUNDARY:
       {
-      BOOL prev_is_word = (eptr != md->start_subject) &&
-        ((md->ctypes[eptr[-1]] & ctype_word) != 0);
+      /* Modified the next line to check previous character
+	 in case we're not at the beginning of the whole string
+	 (Andrey Zmievski) */
+      BOOL prev_is_word = (eptr != md->start_subject) ?
+        ((md->ctypes[eptr[-1]] & ctype_word) != 0) :
+		((md->ctypes[md->regprev] & ctype_word) != 0);
       BOOL cur_is_word = (eptr < md->end_subject) &&
         ((md->ctypes[*eptr] & ctype_word) != 0);
       if ((*ecode++ == OP_WORD_BOUNDARY)?
@@ -4125,7 +4129,8 @@ Returns:          > 0 => success; value is the number of elements filled in
 
 int
 pcre_exec(const pcre *external_re, const pcre_extra *external_extra,
-  const char *subject, int length, int options, int *offsets, int offsetcount, int minlen)
+  const char *subject, int length, const char *strbeg, int options,
+  int *offsets, int offsetcount, int minlen)
 {
 int resetcount, ocount;
 int first_char = -1;
@@ -4160,6 +4165,15 @@ match_block.errorcode = PCRE_ERROR_NOMATCH;     /* Default error */
 match_block.lcc = re->tables + lcc_offset;
 match_block.ctypes = re->tables + ctypes_offset;
 
+/* Setup previous character (Andrey Zmievski) */
+if (subject == strbeg)
+	match_block.regprev = '\n';
+else {
+	match_block.regprev = subject[-1];
+	if (!(re->options & PCRE_MULTILINE) && match_block.regprev == '\n')
+		match_block.regprev = '\0';
+}
+
 /* The ims options can vary during the matching as a result of the presence
 of (?ims) items in the pattern. They are kept in a local variable so that
 restoring at the exit of a group is easy. */
diff --git a/ext/pcre/pcrelib/pcre.h b/ext/pcre/pcrelib/pcre.h
index 0efc0f6502..8fedb8189e 100644
--- a/ext/pcre/pcrelib/pcre.h
+++ b/ext/pcre/pcrelib/pcre.h
@@ -59,7 +59,7 @@ extern pcre *pcre_compile(const char *, int, const char **, int *,
   const unsigned char *);
 extern int pcre_copy_substring(const char *, int *, int, int, char *, int);
 extern int pcre_exec(const pcre *, const pcre_extra *, const char *,
-  int, int, int *, int, int);
+  int, const char *, int, int *, int, int);
 extern int pcre_get_substring(const char *, int *, int, int, const char **);
 extern int pcre_get_substring_list(const char *, int *, int, const char ***);
 extern int pcre_info(const pcre *, int *, int *);