Fix zero-length matches for multibyte characters

src/nre.nim

@@ -352,16 +352,16 @@ iterator findIter*(str: string, pattern: Regex, start = 0, endpos = int.high): R
     if match.isNone:
       # either the end of the input or the string
       # cannot be split here
-      offset += 1
-
       if matchesCrLf and offset < (str.len - 1) and
          str[offset] == '\r' and str[offset + 1] == '\l':
         # if PCRE treats CrLf as newline, skip both at the same time
-        offset += 1
+        offset += 2
       elif unicode:
         # XXX what about invalid unicode?
         offset += str.runeLenAt(offset)
         assert(offset <= strlen)
+      else:
+        offset += 1
     else:
       offset = match.get.matchBounds.b + 1
 

test/find.nim

@@ -19,4 +19,6 @@ suite "find":
   test "len 0 find":
     check("".findAll(re"\ ") == newSeq[string]())
     check("".findAll(re"") == @[""])
-    check("word word".findAll(nre.re"\b") == @["", "", "", ""])
+    check("word word".findAll(re"\b") == @["", "", "", ""])
+    check("word\r\lword".findAll(re(r"$", "m<anycrlf>")) == @["", ""])
+    check("слово слово".findAll(re(r"\b", "uW")) == @["", "", "", ""])

test/misc.nim

@@ -4,4 +4,3 @@ suite "Misc tests":
   test "unicode":
     check("".find(re("", "8")).match == "")
     check("перевірка".replace(re(r"\w", "uW"), "") == "")
-

test/split.nim

@@ -21,6 +21,8 @@ suite "string splitting":
     check("12345".split(re("")) == @["1", "2", "3", "4", "5"])
     check("".split(re"") == newSeq[string]())
     check("word word".split(re"\b") == @["word", " ", "word"])
+    check("word\r\lword".split(re(r"$", "m<anycrlf>")) == @["word", "\r\lword"])
+    check("слово слово".split(re(r"(\b)", "uW")) == @["", "слово", "", " ", "", "слово", ""])
 
   test "perl split tests":
     check("forty-two"                    .split(re"")      .join(",") == "f,o,r,t,y,-,t,w,o")