Shrink isValidNameNonASCII by using U16_NEXT only once.

Source/core/dom/Document.cpp

Index: Source/core/dom/Document.cpp
diff --git a/Source/core/dom/Document.cpp b/Source/core/dom/Document.cpp
index 4b97db5687594905c5098d97a830f72460741aa7..2c24d0f76f36da5bbe38fc421f3cecb33840e009 100644
--- a/Source/core/dom/Document.cpp
+++ b/Source/core/dom/Document.cpp
@@ -4105,16 +4105,11 @@ static bool isValidNameNonASCII(const LChar* characters, unsigned length)
 
 static bool isValidNameNonASCII(const UChar* characters, unsigned length)
 {
-    unsigned i = 0;
-
-    UChar32 c;
-    U16_NEXT(characters, i, length, c)
-    if (!isValidNameStart(c))
-        return false;
-
-    while (i < length) {
-        U16_NEXT(characters, i, length, c)
-        if (!isValidNamePart(c))
+    for (unsigned i = 0; i < length;) {

[fs, 2014/09/04 12:01:18]

Some might be alarmed by a for lacking an update expression - but I see
parseQualifiedNameInternal does it like this (and it's not mentioned in coding
style), so maybe it's ok...

Aside: Speaking of parseQualifiedName... that seems to be quite the beast as
well (almost 5k!). Sinking the exception/error-reporting out would probably put
a significant dent into that...

+        bool first = i == 0;

[fs, 2014/09/04 12:01:18]

I suppose this could also be:

bool nameStart = true;
for (...) {
    if (nameStart ? ...)
        ...;
    nameStart = false;
}

Hard to say what's better codegen-wise though...

+        UChar32 c;
+        U16_NEXT(characters, i, length, c); // Increments i.
+        if (first ? !isValidNameStart(c) : !isValidNamePart(c))
             return false;
     }