Bug 534136 Part 2: Use 16bit-char buffers inside atoms. r=bz sr=mrbkap

accessible/src/base/nsAccessible.cpp

@@ -1491,9 +1491,8 @@ nsAccessible::GetAttributes(nsIPersistentProperties **aAttributes)
     const nsAttrName *attr = content->GetAttrNameAt(count);
     if (attr && attr->NamespaceEquals(kNameSpaceID_None)) {
       nsIAtom *attrAtom = attr->Atom();
-      const char *attrStr;
+      nsDependentAtomString attrStr(attrAtom);
-      attrAtom->GetUTF8String(&attrStr);
+      if (!StringBeginsWith(attrStr, NS_LITERAL_STRING("aria-"))) 
-      if (PL_strncmp(attrStr, "aria-", 5)) 
         continue; // Not ARIA
       PRUint8 attrFlags = nsAccUtils::GetAttributeCharacteristics(attrAtom);
       if (attrFlags & ATTR_BYPASSOBJ)
@@ -1503,7 +1502,7 @@ nsAccessible::GetAttributes(nsIPersistentProperties **aAttributes)
         continue; // only expose token based attributes if they are defined
       nsAutoString value;
       if (content->GetAttr(kNameSpaceID_None, attrAtom, value)) {
-        attributes->SetStringProperty(nsDependentCString(attrStr + 5), value, oldValueUnused);
+        attributes->SetStringProperty(NS_ConvertUTF16toUTF8(Substring(attrStr, 5)), value, oldValueUnused);
       }
     }
   }

accessible/src/base/nsDocAccessible.cpp

@@ -1112,9 +1112,8 @@ nsDocAccessible::AttributeChangedImpl(nsIContent* aContent, PRInt32 aNameSpaceID
   // Check for namespaced ARIA attribute
   if (aNameSpaceID == kNameSpaceID_None) {
     // Check for hyphenated aria-foo property?
-    const char* attributeName;
+    if (StringBeginsWith(nsDependentAtomString(aAttribute),
-    aAttribute->GetUTF8String(&attributeName);
+                         NS_LITERAL_STRING("aria-"))) {
-    if (!PL_strncmp("aria-", attributeName, 5)) {
       ARIAAttributeChanged(aContent, aAttribute);
     }
   }

accessible/src/msaa/nsAccessNodeWrap.cpp

@@ -249,12 +249,10 @@ __try{
   for (PRUint32 index = 0; index < numAttribs; index++) {
     aNameSpaceIDs[index] = 0; aAttribValues[index] = aAttribNames[index] = nsnull;
     nsAutoString attributeValue;
-    const char *pszAttributeName; 
 
     const nsAttrName* name = content->GetAttrNameAt(index);
     aNameSpaceIDs[index] = static_cast<short>(name->NamespaceID());
-    name->LocalName()->GetUTF8String(&pszAttributeName);
+    aAttribNames[index] = ::SysAllocString(name->LocalName()->GetUTF16String());
-    aAttribNames[index] = ::SysAllocString(NS_ConvertUTF8toUTF16(pszAttributeName).get());
     content->GetAttr(name->NamespaceID(), name->LocalName(), attributeValue);
     aAttribValues[index] = ::SysAllocString(attributeValue.get());
   }

content/base/public/nsContentUtils.h

@@ -1455,6 +1455,13 @@ public:
 
   static JSContext *GetCurrentJSContext();
 
+  /**
+   * Case insensitive comparison between two strings. However it only ignores
+   * case for ASCII characters a-z.
+   */
+  static PRBool EqualsIgnoreASCIICase(const nsAString& aStr1,
+                                      const nsAString& aStr2);
+
   /**
    * Convert ASCII A-Z to a-z.
    */

content/base/public/nsINodeInfo.h

@@ -270,9 +270,7 @@ public:
     if (!GetPrefixAtom())
       return Equals(aNameAtom);
 
-    const char* utf8;
+    return QualifiedNameEqualsInternal(nsAtomCString(aNameAtom));
-    aNameAtom->GetUTF8String(&utf8);
-    return QualifiedNameEqualsInternal(nsDependentCString(utf8));
   }
 
   PRBool QualifiedNameEquals(const nsACString& aQualifiedName) const

content/base/src/nsAttrValue.cpp

@@ -510,7 +510,7 @@ nsAttrValue::HashValue() const
       nsStringBuffer* str = static_cast<nsStringBuffer*>(GetPtr());
       if (str) {
         PRUint32 len = str->StorageSize()/sizeof(PRUnichar) - 1;
-        return nsCRT::BufferHashCode(static_cast<PRUnichar*>(str->Data()), len);
+        return nsCRT::HashCode(static_cast<PRUnichar*>(str->Data()), len);
       }
 
       return 0;
@@ -776,11 +776,11 @@ nsAttrValue::Contains(nsIAtom* aValue, nsCaseTreatment aCaseSensitive) const
         return aValue == atom;
       }
 
-      const char *val1, *val2;
+      // For performance reasons, don't do a full on unicode case insensitive
-      aValue->GetUTF8String(&val1);
+      // string comparison. This is only used for quirks mode anyway.
-      atom->GetUTF8String(&val2);
+      return
-
+        nsContentUtils::EqualsIgnoreASCIICase(nsDependentAtomString(aValue),
-      return nsCRT::strcasecmp(val1, val2) == 0;
+                                              nsDependentAtomString(atom));
     }
     default:
     {
@@ -790,12 +790,14 @@ nsAttrValue::Contains(nsIAtom* aValue, nsCaseTreatment aCaseSensitive) const
           return array->IndexOf(aValue) >= 0;
         }
 
-        const char *val1, *val2;
+        nsDependentAtomString val1(aValue);
-        aValue->GetUTF8String(&val1);
 
         for (PRInt32 i = 0, count = array->Count(); i < count; ++i) {
-          array->ObjectAt(i)->GetUTF8String(&val2);
+          // For performance reasons, don't do a full on unicode case
-          if (nsCRT::strcasecmp(val1, val2) == 0) {
+          // insensitive string comparison. This is only used for quirks mode
+          // anyway.
+          if (nsContentUtils::EqualsIgnoreASCIICase(val1,
+                nsDependentAtomString(array->ObjectAt(i)))) {
             return PR_TRUE;
           }
         }

content/base/src/nsContentSink.cpp

@@ -532,9 +532,7 @@ nsContentSink::ProcessHeaderData(nsIAtom* aHeader, const nsAString& aValue,
     if (NS_SUCCEEDED(mParser->GetChannel(getter_AddRefs(channel)))) {
       nsCOMPtr<nsIHttpChannel> httpChannel(do_QueryInterface(channel));
       if (httpChannel) {
-        const char* header;
+        httpChannel->SetResponseHeader(nsAtomCString(aHeader),
-        (void)aHeader->GetUTF8String(&header);
-        (void)httpChannel->SetResponseHeader(nsDependentCString(header),
                                        NS_ConvertUTF16toUTF8(aValue),
                                        PR_TRUE);
       }

content/base/src/nsContentUtils.cpp

@@ -1897,10 +1897,7 @@ static inline void KeyAppendAtom(nsIAtom* aAtom, nsACString& aKey)
 {
   NS_PRECONDITION(aAtom, "KeyAppendAtom: aAtom can not be null!\n");
 
-  const char* atomString = nsnull;
+  KeyAppendString(nsAtomCString(aAtom), aKey);
-  aAtom->GetUTF8String(&atomString);
-
-  KeyAppendString(nsDependentCString(atomString), aKey);
 }
 
 static inline PRBool IsAutocompleteOff(nsIDOMElement* aElement)
@@ -3210,8 +3207,7 @@ nsAutoGCRoot::RemoveJSGCRoot(void* aPtr)
 PRBool
 nsContentUtils::IsEventAttributeName(nsIAtom* aName, PRInt32 aType)
 {
-  const char* name;
+  const PRUnichar* name = aName->GetUTF16String();
-  aName->GetUTF8String(&name);
   if (name[0] != 'o' || name[1] != 'n')
     return PR_FALSE;
 
@@ -4930,6 +4926,43 @@ nsContentUtils::ASCIIToUpper(nsAString& aStr)
   }
 }
 
+PRBool
+nsContentUtils::EqualsIgnoreASCIICase(const nsAString& aStr1,
+                                      const nsAString& aStr2)
+{
+  PRUint32 len = aStr1.Length();
+  if (len != aStr2.Length()) {
+    return PR_FALSE;
+  }
+
+  const PRUnichar* str1 = aStr1.BeginReading();
+  const PRUnichar* str2 = aStr2.BeginReading();
+  const PRUnichar* end = str1 + len;
+
+  while (str1 < end) {
+    PRUnichar c1 = *str1++;
+    PRUnichar c2 = *str2++;
+
+    // First check if any bits other than the 0x0020 differs
+    if ((c1 ^ c2) & 0xffdf) {
+      return PR_FALSE;
+    }
+
+    // We know they only differ in the 0x0020 bit.
+    // Likely the two chars are the same, so check that first
+    if (c1 != c2) {
+      // They do differ, but since it's only in the 0x0020 bit, check if it's
+      // the same ascii char, but just differing in case
+      PRUnichar c1Upper = c1 & 0xffdf;
+      if (!('A' <= c1Upper && c1Upper <= 'Z')) {
+        return PR_FALSE;
+      }
+    }
+  }
+
+  return PR_TRUE;
+}
+
 /* static */
 void
 nsAutoGCRoot::Shutdown()

content/base/src/nsDocument.cpp

@@ -6718,9 +6718,7 @@ nsDocument::CreateElem(nsIAtom *aName, nsIAtom *aPrefix, PRInt32 aNamespaceID,
     aPrefix->ToString(qName);
     qName.Append(':');
   }
-  const char *name;
+  qName.Append(nsAtomString(aName));
-  aName->GetUTF8String(&name);
-  AppendUTF8toUTF16(name, qName);
 
   // Note: "a:b:c" is a valid name in non-namespaces XML, and
   // nsDocument::CreateElement can call us with such a name and no prefix,

content/base/src/nsHTMLContentSerializer.cpp

@@ -69,8 +69,6 @@
 #include "nsIScriptElement.h"
 #include "nsAttrName.h"
 
-static const char kMozStr[] = "moz";
-
 static const PRInt32 kLongLineLen = 128;
 
 nsresult NS_NewHTMLContentSerializer(nsIContentSerializer** aSerializer)
@@ -123,10 +121,9 @@ nsHTMLContentSerializer::SerializeHTMLAttributes(nsIContent* aContent,
     nsIAtom* attrName = name->LocalName();
 
     // Filter out any attribute starting with [-|_]moz
-    const char* sharedName;
+    nsDependentAtomString attrNameStr(attrName);
-    attrName->GetUTF8String(&sharedName);
+    if (StringBeginsWith(attrNameStr, NS_LITERAL_STRING("_moz")) ||
-    if ((('_' == *sharedName) || ('-' == *sharedName)) &&
+        StringBeginsWith(attrNameStr, NS_LITERAL_STRING("-moz"))) {
-        !nsCRT::strncmp(sharedName+1, kMozStr, PRUint32(sizeof(kMozStr)-1))) {
       continue;
     }
     aContent->GetAttr(namespaceID, attrName, valueStr);

content/base/src/nsNodeInfo.cpp

@@ -257,16 +257,13 @@ nsNodeInfo::QualifiedNameEqualsInternal(const nsACString& aQualifiedName) const
 
   nsACString::const_iterator colon(start);
 
-  const char* prefix;
+  nsAtomCString prefix(mInner.mPrefix);
-  mInner.mPrefix->GetUTF8String(&prefix);
 
-  PRUint32 len = strlen(prefix);
+  if (prefix.Length() >= aQualifiedName.Length()) {
-
-  if (len >= aQualifiedName.Length()) {
     return PR_FALSE;
   }
 
-  colon.advance(len);
+  colon.advance(prefix.Length());
 
   // If the character at the prefix length index is not a colon,
   // aQualifiedName is not equal to this string.
@@ -275,7 +272,7 @@ nsNodeInfo::QualifiedNameEqualsInternal(const nsACString& aQualifiedName) const
   }
 
   // Compare the prefix to the string from the start to the colon
-  if (!mInner.mPrefix->EqualsUTF8(Substring(start, colon)))
+  if (!prefix.Equals(Substring(start, colon)))
     return PR_FALSE;
 
   ++colon; // Skip the ':'

content/base/src/nsXHTMLContentSerializer.cpp

@@ -346,10 +346,9 @@ nsXHTMLContentSerializer::SerializeAttributes(nsIContent* aContent,
     nsIAtom* attrPrefix = name->GetPrefix();
 
     // Filter out any attribute starting with [-|_]moz
-    const char* sharedName;
+    nsDependentAtomString attrNameStr(attrName);
-    attrName->GetUTF8String(&sharedName);
+    if (StringBeginsWith(attrNameStr, NS_LITERAL_STRING("_moz")) ||
-    if ((('_' == *sharedName) || ('-' == *sharedName)) &&
+        StringBeginsWith(attrNameStr, NS_LITERAL_STRING("-moz"))) {
-        !nsCRT::strncmp(sharedName+1, kMozStr, PRUint32(sizeof(kMozStr)-1))) {
       continue;
     }
 

content/html/document/src/nsHTMLContentSink.cpp

@@ -942,13 +942,10 @@ SinkContext::CloseContainer(const nsHTMLTag aTag, PRBool aMalformed)
 #ifdef NS_DEBUG
       {
         // Tracing code
-        const char *tagStr;
-        mStack[mStackPos].mContent->Tag()->GetUTF8String(&tagStr);
-
         SINK_TRACE(gSinkLogModuleInfo, SINK_TRACE_REFLOW,
                    ("SinkContext::CloseContainer: reflow on notifyImmediate "
                     "tag=%s newIndex=%d stackPos=%d",
-                    tagStr,
+                    nsAtomCString(mStack[mStackPos].mContent->Tag()).get(),
                     mStack[mStackPos].mNumFlushed, mStackPos));
       }
 #endif
@@ -1353,12 +1350,10 @@ SinkContext::FlushTags()
 #ifdef NS_DEBUG
         {
           // Tracing code
-          const char* tagStr;
-          mStack[stackPos].mContent->Tag()->GetUTF8String(&tagStr);
-
           SINK_TRACE(gSinkLogModuleInfo, SINK_TRACE_REFLOW,
                      ("SinkContext::FlushTags: tag=%s from newindex=%d at "
-                      "stackPos=%d", tagStr,
+                      "stackPos=%d",
+                      nsAtomCString(mStack[stackPos].mContent->Tag()).get(),
                       mStack[stackPos].mNumFlushed, stackPos));
         }
 #endif

content/smil/nsSMILCompositor.cpp

@@ -51,10 +51,8 @@ nsSMILCompositor::KeyEquals(KeyTypePointer aKey) const
 nsSMILCompositor::HashKey(KeyTypePointer aKey)
 {
   // Combine the 3 values into one numeric value, which will be hashed
-  const char *attrName = nsnull;
+  return NS_PTR_TO_UINT32(aKey->mElement.get()) >> 4 +
-  aKey->mAttributeName->GetUTF8String(&attrName);
+    NS_PTR_TO_INT32(aKey->mAttributeName.get()) +
-  return NS_PTR_TO_UINT32(aKey->mElement.get()) +
-    HashString(attrName) +
     (aKey->mIsCSS ? 1 : 0);
 }
 

content/xslt/src/base/txStringUtils.h

@@ -49,10 +49,7 @@
 inline PRBool
 TX_StringEqualsAtom(const nsASingleFragmentString& aString, nsIAtom* aAtom)
 {
-    const char* ASCIIAtom;
+  return aAtom->Equals(aString);
-    aAtom->GetUTF8String(&ASCIIAtom);
-
-    return aString.EqualsASCII(ASCIIAtom);
 }
 
 #ifndef TX_EXE

content/xslt/src/xpath/txXPCOMExtensionFunction.cpp

@@ -212,10 +212,9 @@ LookupFunction(const char *aContractID, nsIAtom* aName, nsIID &aIID,
     // foo-bar becomes fooBar). Note that if there are any names that already
     // have uppercase letters they might cause false matches (both fooBar and
     // foo-bar matching fooBar).
-    const char *name;
+    const PRUnichar *name = aName->GetUTF16String();
-    aName->GetUTF8String(&name);
     nsCAutoString methodName;
-    char letter;
+    PRUnichar letter;
     PRBool upperNext = PR_FALSE;
     while ((letter = *name)) {
         if (letter == '-') {

content/xul/document/src/nsXULDocument.cpp

@@ -1408,20 +1408,17 @@ nsXULDocument::Persist(nsIContent* aElement, PRInt32 aNameSpaceID,
 
     // Ick. Construct a property from the attribute. Punt on
     // namespaces for now.
-    const char* attrstr;
-    rv = aAttribute->GetUTF8String(&attrstr);
-    if (NS_FAILED(rv)) return rv;
-
     // Don't bother with unreasonable attributes. We clamp long values,
     // but truncating attribute names turns it into a different attribute
     // so there's no point in persisting anything at all
-    if (!attrstr || strlen(attrstr) > kMaxAttrNameLength) {
+    nsAtomCString attrstr(aAttribute);
+    if (attrstr.Length() > kMaxAttrNameLength) {
         NS_WARNING("Can't persist, Attribute name too long");
         return NS_ERROR_ILLEGAL_VALUE;
     }
 
     nsCOMPtr<nsIRDFResource> attr;
-    rv = gRDFService->GetResource(nsDependentCString(attrstr),
+    rv = gRDFService->GetResource(attrstr,
                                   getter_AddRefs(attr));
     if (NS_FAILED(rv)) return rv;
 

content/xul/templates/src/nsXULContentBuilder.cpp

@@ -483,17 +483,14 @@ nsXULContentBuilder::BuildContentFromTemplate(nsIContent *aTemplateNode,
                ("nsXULContentBuilder::BuildContentFromTemplate (is unique: %d)",
                aIsUnique));
 
-        const char *tmpln, *resn, *realn;
-        aTemplateNode->Tag()->GetUTF8String(&tmpln);
-        aResourceNode->Tag()->GetUTF8String(&resn);
-        aRealNode->Tag()->GetUTF8String(&realn);
-
         nsAutoString id;
         aChild->GetId(id);
 
         PR_LOG(gXULTemplateLog, PR_LOG_ALWAYS,
                ("Tags: [Template: %s  Resource: %s  Real: %s] for id %s",
-               tmpln, resn, realn, NS_ConvertUTF16toUTF8(id).get()));
+                nsAtomCString(aTemplateNode->Tag()).get(), 
+                nsAtomCString(aResourceNode->Tag()).get(),
+                nsAtomCString(aRealNode->Tag()).get(), NS_ConvertUTF16toUTF8(id).get()));
     }
 #endif
 
@@ -562,11 +559,9 @@ nsXULContentBuilder::BuildContentFromTemplate(nsIContent *aTemplateNode,
 
 #ifdef PR_LOGGING
         if (PR_LOG_TEST(gXULTemplateLog, PR_LOG_DEBUG)) {
-            const char *tagname;
-            tag->GetUTF8String(&tagname);
             PR_LOG(gXULTemplateLog, PR_LOG_DEBUG,
                    ("xultemplate[%p]     building %s %s %s",
-                    this, tagname,
+                    this, nsAtomCString(tag).get(),
                     (isGenerationElement ? "[resource]" : ""),
                     (isUnique ? "[unique]" : "")));
         }

dom/base/nsJSEnvironment.cpp

@@ -1927,26 +1927,24 @@ nsJSContext::ExecuteScript(void *aScriptObject,
 }
 
 
-static inline const char *
-AtomToEventHandlerName(nsIAtom *aName)
-{
-  const char *name;
-
-  aName->GetUTF8String(&name);
-
 #ifdef DEBUG
-  const char *cp;
+PRBool
-  char c;
+AtomIsEventHandlerName(nsIAtom *aName)
+{
+  const PRUnichar *name = aName->GetUTF16String();
+
+  const PRUnichar *cp;
+  PRUnichar c;
   for (cp = name; *cp != '\0'; ++cp)
   {
     c = *cp;
-    NS_ASSERTION (('A' <= c && c <= 'Z') || ('a' <= c && c <= 'z'),
+    if ((c < 'A' || c > 'Z') && (c < 'a' || c > 'z'))
-                  "non-ASCII non-alphabetic event handler name");
+      return PR_FALSE;
   }
-#endif
 
-  return name;
+  return PR_TRUE;
 }
+#endif
 
 // Helper function to find the JSObject associated with a (presumably DOM)
 // interface.
@@ -1992,6 +1990,7 @@ nsJSContext::CompileEventHandler(nsIAtom *aName,
 {
   NS_ENSURE_TRUE(mIsInitialized, NS_ERROR_NOT_INITIALIZED);
 
+  NS_PRECONDITION(AtomIsEventHandlerName(aName), "Bad event name");
   NS_PRECONDITION(!::JS_IsExceptionPending(mContext),
                   "Why are we being called with a pending exception?");
 
@@ -2008,8 +2007,6 @@ nsJSContext::CompileEventHandler(nsIAtom *aName,
     return NS_ERROR_ILLEGAL_VALUE;
   }
 
-  const char *charName = AtomToEventHandlerName(aName);
-
 #ifdef DEBUG
   JSContext* top = nsContentUtils::GetCurrentJSContext();
   NS_ASSERTION(mContext == top, "Context not properly pushed!");
@@ -2024,7 +2021,7 @@ nsJSContext::CompileEventHandler(nsIAtom *aName,
   JSFunction* fun =
       ::JS_CompileUCFunctionForPrincipals(mContext,
                                           nsnull, nsnull,
-                                          charName, aArgCount, aArgNames,
+                                          nsAtomCString(aName).get(), aArgCount, aArgNames,
                                           (jschar*)PromiseFlatString(aBody).get(),
                                           aBody.Length(),
                                           aURL, aLineNo);
@@ -2206,7 +2203,7 @@ nsJSContext::BindCompiledEventHandler(nsISupports* aTarget, void *aScope,
   NS_ENSURE_ARG(aHandler);
   NS_ENSURE_TRUE(mIsInitialized, NS_ERROR_NOT_INITIALIZED);
 
-  const char *charName = AtomToEventHandlerName(aName);
+  NS_PRECONDITION(AtomIsEventHandlerName(aName), "Bad event name");
   nsresult rv;
 
   // Get the jsobject associated with this target
@@ -2240,7 +2237,7 @@ nsJSContext::BindCompiledEventHandler(nsISupports* aTarget, void *aScope,
 
   if (NS_SUCCEEDED(rv) &&
       // Make sure the flags here match those in nsEventReceiverSH::NewResolve
-      !::JS_DefineProperty(mContext, target, charName,
+      !::JS_DefineProperty(mContext, target, nsAtomCString(aName).get(),
                            OBJECT_TO_JSVAL(funobj), nsnull, nsnull,
                            JSPROP_ENUMERATE | JSPROP_PERMANENT)) {
     ReportPendingException();
@@ -2262,6 +2259,8 @@ nsJSContext::GetBoundEventHandler(nsISupports* aTarget, void *aScope,
                                   nsIAtom* aName,
                                   nsScriptObjectHolder &aHandler)
 {
+    NS_PRECONDITION(AtomIsEventHandlerName(aName), "Bad event name");
+
     nsresult rv;
     JSObject *obj = nsnull;
     nsAutoGCRoot root(&obj, &rv);
@@ -2270,11 +2269,9 @@ nsJSContext::GetBoundEventHandler(nsISupports* aTarget, void *aScope,
     rv = JSObjectFromInterface(aTarget, aScope, &obj);
     NS_ENSURE_SUCCESS(rv, rv);
 
-    const char *charName = AtomToEventHandlerName(aName);
-
     jsval funval;
     if (!JS_LookupProperty(mContext, obj,
-                           charName, &funval))
+                           nsAtomCString(aName).get(), &funval))
         return NS_ERROR_FAILURE;
 
     if (JS_TypeOfValue(mContext, funval) != JSTYPE_FUNCTION) {

gfx/thebes/src/gfxWindowsFonts.cpp

@@ -1860,9 +1860,7 @@ gfxWindowsFontGroup::WhichPrefFontSupportsChar(PRUint32 aCh)
                 nsIAtom *langGroup = LangGroupFromUnicodeRange(unicodeRange);
                 if (langGroup) {
 #ifdef PR_LOGGING
-                    const char *langGroupStr;
+                    PR_LOG(gFontLog, PR_LOG_DEBUG, (" - Trying to find fonts for: %s", nsAtomCString(langGroup).get()));
-                    langGroup->GetUTF8String(&langGroupStr);
-                    PR_LOG(gFontLog, PR_LOG_DEBUG, (" - Trying to find fonts for: %s", langGroupStr));
 #endif
                     nsAutoTArray<nsRefPtr<gfxFontEntry>, 5> fonts;
                     this->GetPrefFonts(langGroup, fonts);

gfx/thebes/test/gfxFontSelectionTests.h

@@ -109,7 +109,7 @@ SetupTests()
                                           NS_FONT_STRETCH_NORMAL,
                                           400,
                                           16.0,
-                                          NS_NewPermanentAtom(NS_LITERAL_CSTRING("en")),
+                                          NS_NewPermanentAtom(NS_LITERAL_STRING("en")),
                                           0.0,
                                           PR_FALSE, PR_FALSE, PR_FALSE);
 
@@ -117,7 +117,7 @@ SetupTests()
                                         NS_FONT_STRETCH_NORMAL,
                                         700,
                                         16.0,
-                                        NS_NewPermanentAtom(NS_LITERAL_CSTRING("en")),
+                                        NS_NewPermanentAtom(NS_LITERAL_STRING("en")),
                                         0.0,
                                         PR_FALSE, PR_FALSE, PR_FALSE);
 

gfx/thebes/test/gfxTextRunPerfTest.cpp

@@ -93,7 +93,7 @@ RunTest (TestEntry *test, gfxContext *ctx) {
                                               NS_FONT_STRETCH_NORMAL,
                                               400,
                                               16.0,
-                                              NS_NewPermanentAtom(NS_LITERAL_CSTRING("en")),
+                                              NS_NewPermanentAtom(NS_LITERAL_STRING("en")),
                                               0.0,
                                               PR_FALSE, PR_FALSE, PR_FALSE);
 

gfx/thebes/test/gfxWordCacheTest.cpp

@@ -159,7 +159,7 @@ main (int argc, char **argv) {
                            NS_FONT_STRETCH_NORMAL,
                            139,
                            10.0,
-                           NS_NewPermanentAtom(NS_LITERAL_CSTRING("en")),
+                           NS_NewPermanentAtom(NS_LITERAL_STRING("en")),
                            0.0,
                            PR_FALSE, PR_FALSE, PR_FALSE);
 

layout/base/nsPresContext.cpp

@@ -431,13 +431,19 @@ nsPresContext::GetFontPreferences()
   mDefaultFixedFont.size = CSSPixelsToAppUnits(13);
 
   // the font prefs are based on langGroup, not actual language
-  const char *langGroup = "x-western"; // Assume x-western is safe...
+  nsCAutoString langGroup;
   if (mLanguage && mLangService) {
     nsresult rv;
     nsIAtom *group = mLangService->GetLanguageGroup(mLanguage, &rv);
     if (NS_SUCCEEDED(rv) && group) {
-      group->GetUTF8String(&langGroup);
+      group->ToUTF8String(langGroup);
     }
+    else {
+      langGroup.AssignLiteral("x-western"); // Assume x-western is safe...
+    }
+  }
+  else {
+    langGroup.AssignLiteral("x-western"); // Assume x-western is safe...
   }
 
   nsCAutoString pref;

layout/generic/nsFrame.cpp

@@ -6734,11 +6734,13 @@ static void
 GetTagName(nsFrame* aFrame, nsIContent* aContent, PRIntn aResultSize,
            char* aResult)
 {
-  const char *nameStr = "";
   if (aContent) {
-    aContent->Tag()->GetUTF8String(&nameStr);
+    PR_snprintf(aResult, aResultSize, "%s@%p",
+                nsAtomCString(aContent->Tag()).get(), aFrame);
+  }
+  else {
+    PR_snprintf(aResult, aResultSize, "@%p", aFrame);
   }
-  PR_snprintf(aResult, aResultSize, "%s@%p", nameStr, aFrame);
 }
 
 void

layout/generic/nsSelection.cpp

@@ -4881,9 +4881,8 @@ nsTypedSelection::Collapse(nsINode* aParentNode, PRInt32 aOffset)
     if (!content)
       return NS_ERROR_FAILURE;
 
-    const char *tagString;
+    printf ("Sel. Collapse to %p %s %d\n", content.get(),
-    content->Tag()->GetUTF8String(&tagString);
+            nsAtomCString(content->Tag()).get(), aOffset);
-    printf ("Sel. Collapse to %p %s %d\n", content.get(), tagString, aOffset);
   }
   else {
     printf ("Sel. Collapse set to null parent.\n");
@@ -5274,9 +5273,8 @@ nsTypedSelection::Extend(nsINode* aParentNode, PRInt32 aOffset)
     nsCOMPtr<nsIContent>content;
     content = do_QueryInterface(aParentNode);
 
-    const char *tagString;
+    printf ("Sel. Extend to %p %s %d\n", content.get(),
-    content->Tag()->GetUTF8String(&tagString);
+            nsAtomCString(content->Tag()).get(), aOffset);
-    printf ("Sel. Extend to %p %s %d\n", content.get(), tagString, aOffset);
   }
   else {
     printf ("Sel. Extend set to null parent.\n");

layout/generic/nsTextFrameThebes.cpp

@@ -2164,10 +2164,10 @@ static PRBool IsChineseOrJapanese(nsIFrame* aFrame)
   if (!language) {
     return PR_FALSE;
   }
-  const char *lang;
+  const PRUnichar *lang = language->GetUTF16String();
-  language->GetUTF8String(&lang);
+  return (!nsCRT::strncmp(lang, NS_LITERAL_STRING("ja").get(), 2) ||
-  return (!strncmp(lang, "ja", 2) || !strncmp(lang, "zh", 2)) &&
+          !nsCRT::strncmp(lang, NS_LITERAL_STRING("zh").get(), 2)) &&
-         (lang[2] == '\0' || lang[2] == '-');
+         (language->GetLength() == 2 || lang[2] == '-');
 }
 
 #ifdef DEBUG

layout/style/nsCSSAnonBoxes.cpp

@@ -78,9 +78,7 @@ PRBool nsCSSAnonBoxes::IsAnonBox(nsIAtom *aAtom)
 /* static */ PRBool
 nsCSSAnonBoxes::IsTreePseudoElement(nsIAtom* aPseudo)
 {
-  const char* str;
+  return StringBeginsWith(nsDependentAtomString(aPseudo),
-  aPseudo->GetUTF8String(&str);
+                          NS_LITERAL_STRING(":-moz-tree-"));
-  static const char moz_tree[] = ":-moz-tree-";
-  return nsCRT::strncmp(str, moz_tree, PRInt32(sizeof(moz_tree)-1)) == 0;
 }
 #endif

layout/style/nsCSSRuleProcessor.cpp

@@ -196,11 +196,13 @@ RuleHash_CIMatchEntry(PLDHashTable *table, const PLDHashEntryHdr *hdr,
   if (match_atom == entry_atom)
     return PR_TRUE;
 
-  const char *match_str, *entry_str;
+  // Use EqualsIgnoreASCIICase instead of full on unicode case conversion
-  match_atom->GetUTF8String(&match_str);
+  // in order to save on performance. This is only used in quirks mode
-  entry_atom->GetUTF8String(&entry_str);
+  // anyway.
 
-  return (nsCRT::strcasecmp(entry_str, match_str) == 0);
+  return
+    nsContentUtils::EqualsIgnoreASCIICase(nsDependentAtomString(entry_atom),
+                                          nsDependentAtomString(match_atom));
 }
 
 static PRBool
@@ -1834,13 +1836,13 @@ static PRBool SelectorMatches(RuleProcessorData &data,
           IDList = IDList->mNext;
         } while (IDList);
       } else {
-        const char* id1Str;
+        // Use EqualsIgnoreASCIICase instead of full on unicode case conversion
-        data.mContentID->GetUTF8String(&id1Str);
+        // in order to save on performance. This is only used in quirks mode
-        nsDependentCString id1(id1Str);
+        // anyway.
+        nsDependentAtomString id1Str(data.mContentID);
         do {
-          const char* id2Str;
+          if (!nsContentUtils::EqualsIgnoreASCIICase(id1Str,
-          IDList->mAtom->GetUTF8String(&id2Str);
+                 nsDependentAtomString(IDList->mAtom))) {
-          if (!id1.Equals(id2Str, nsCaseInsensitiveCStringComparator())) {
             return PR_FALSE;
           }
           IDList = IDList->mNext;

layout/style/nsComputedDOMStyle.cpp

@@ -4294,10 +4294,9 @@ nsComputedDOMStyle::GetTransitionProperty(nsIDOMCSSValue** aValue)
       property->SetIdent(eCSSKeyword_none);
     else if (cssprop == eCSSProperty_UNKNOWN)
     {
-      const char *str;
-      transition->GetUnknownProperty()->GetUTF8String(&str);
       nsAutoString escaped;
-      nsStyleUtil::AppendEscapedCSSIdent(NS_ConvertUTF8toUTF16(str), escaped);
+      nsStyleUtil::AppendEscapedCSSIdent(
+        nsDependentAtomString(transition->GetUnknownProperty()), escaped);
       property->SetString(escaped); // really want SetIdent
     }
     else

parser/html/nsHtml5Atom.cpp

@@ -82,9 +82,16 @@ nsHtml5Atom::ToUTF8String(nsACString& aReturn)
 }
 
 NS_IMETHODIMP
-nsHtml5Atom::GetUTF8String(const char **aReturn)
+nsHtml5Atom::GetUTF16String(const PRUnichar **aReturn)
 {
-  NS_NOTREACHED("Should not attempt to get a UTF-8 string from nsHtml5Atom");
+  NS_NOTREACHED("Should not attempt to get a UTF-16 string from nsHtml5Atom");
+  return NS_ERROR_NOT_IMPLEMENTED;
+}
+
+NS_IMETHODIMP_(PRUint32)
+nsHtml5Atom::GetLength()
+{
+  NS_NOTREACHED("Should not attempt to get a length from nsHtml5Atom");
   return NS_ERROR_NOT_IMPLEMENTED;
 }
 

parser/htmlparser/src/nsHTMLTags.cpp

@@ -368,12 +368,12 @@ nsHTMLTags::AddRefTable(void)
 
 
 
-#ifdef DEBUG
+#if defined(DEBUG) && defined(NS_STATIC_ATOM_USE_WIDE_STRINGS)
     {
       // let's verify that all names in the the table are lowercase...
       for (i = 0; i < NS_HTML_TAG_MAX; ++i) {
-        nsCAutoString temp1((char*)sTagAtoms_info[i].mStringBuffer->Data());
+        nsAutoString temp1((PRUnichar*)sTagAtoms_info[i].mStringBuffer->Data());
-        nsCAutoString temp2((char*)sTagAtoms_info[i].mStringBuffer->Data());
+        nsAutoString temp2((PRUnichar*)sTagAtoms_info[i].mStringBuffer->Data());
         ToLowerCase(temp1);
         NS_ASSERTION(temp1.Equals(temp2), "upper case char in table");
       }
@@ -382,7 +382,7 @@ nsHTMLTags::AddRefTable(void)
       // correct.
       for (i = 0; i < NS_HTML_TAG_MAX; ++i) {
         nsAutoString temp1(sTagUnicodeTable[i]);
-        nsAutoString temp2; temp2.AssignWithConversion((char*)sTagAtoms_info[i].mStringBuffer->Data());
+        nsAutoString temp2((PRUnichar*)sTagAtoms_info[i].mStringBuffer->Data());
         NS_ASSERTION(temp1.Equals(temp2), "Bad unicode tag name!");
       }
 

rdf/base/src/nsRDFContentSink.cpp

@@ -1026,11 +1026,8 @@ RDFContentSinkImpl::AddProperties(const PRUnichar** aAttributes,
           }
       }
 
-      const char* attrName;
-      localName->GetUTF8String(&attrName);
-
       NS_ConvertUTF16toUTF8 propertyStr(nameSpaceURI);    
-      propertyStr.Append(attrName);
+      propertyStr.Append(nsAtomCString(localName));
 
       // Add the assertion to RDF
       nsCOMPtr<nsIRDFResource> property;
@@ -1158,11 +1155,8 @@ RDFContentSinkImpl::OpenObject(const PRUnichar* aName,
     }
 
     if (isaTypedNode) {
-        const char* attrName;
-        localName->GetUTF8String(&attrName);
-
         NS_ConvertUTF16toUTF8 typeStr(nameSpaceURI);
-        typeStr.Append(attrName);
+        typeStr.Append(nsAtomCString(localName));
 
         nsCOMPtr<nsIRDFResource> type;
         nsresult rv = gRDFService->GetResource(typeStr, getter_AddRefs(type));
@@ -1190,11 +1184,8 @@ RDFContentSinkImpl::OpenProperty(const PRUnichar* aName, const PRUnichar** aAttr
     const nsDependentSubstring& nameSpaceURI =
         SplitExpatName(aName, getter_AddRefs(localName));
 
-    const char* attrName;
-    localName->GetUTF8String(&attrName);
-
     NS_ConvertUTF16toUTF8 propertyStr(nameSpaceURI);
-    propertyStr.Append(attrName);
+    propertyStr.Append(nsAtomCString(localName));
 
     nsCOMPtr<nsIRDFResource> property;
     rv = gRDFService->GetResource(propertyStr, getter_AddRefs(property));

xpcom/ds/nsAtomService.cpp

@@ -59,7 +59,7 @@ nsAtomService::GetAtom(const nsAString& aString, nsIAtom ** aResult)
 nsresult
 nsAtomService::GetPermanentAtom(const nsAString& aString, nsIAtom ** aResult)
 {
-  *aResult = NS_NewPermanentAtom(NS_ConvertUTF16toUTF8(aString));
+  *aResult = NS_NewPermanentAtom(aString);
 
   if (!*aResult)
     return NS_ERROR_OUT_OF_MEMORY;
@@ -81,7 +81,7 @@ nsAtomService::GetAtomUTF8(const char *aValue, nsIAtom* *aResult)
 NS_IMETHODIMP
 nsAtomService::GetPermanentAtomUTF8(const char *aValue, nsIAtom* *aResult)
 {
-    *aResult = NS_NewPermanentAtom(nsDependentCString(aValue));
+    *aResult = NS_NewPermanentAtom(NS_ConvertUTF8toUTF16(aValue));
 
     if (!*aResult)
         return NS_ERROR_OUT_OF_MEMORY;

xpcom/ds/nsAtomTable.cpp

@@ -153,7 +153,7 @@ struct AtomTableEntry : public PLDHashEntryHdr {
   // type-agnostic accessors
 
   // get the string buffer
-  inline const char* getAtomString() const {
+  inline const PRUnichar* getAtomString() const {
     NS_ASSERTION(keyHash > 1,
                  "getAtomString() called on non-atom AtomTableEntry!");
 
@@ -200,14 +200,14 @@ AtomTableGetHash(PLDHashTable *table, const void *key)
 {
   const AtomTableEntry *e = static_cast<const AtomTableEntry*>(key);
 
-  if (e->IsUTF16String()) {
+  if (e->IsUTF8String()) {
-    return nsCRT::HashCodeAsUTF8(e->getUTF16String(), e->getLength());;
+    return nsCRT::HashCodeAsUTF16(e->getUTF8String(), e->getLength());;
   }
 
-  NS_ASSERTION(e->IsUTF8String(),
+  NS_ASSERTION(e->IsUTF16String(),
                "AtomTableGetHash() called on non-string-key AtomTableEntry!");
 
-  return nsCRT::HashCode(e->getUTF8String(), e->getLength());
+  return nsCRT::HashCode(e->getUTF16String(), e->getLength());
 }
 
 static PRBool
@@ -217,13 +217,13 @@ AtomTableMatchKey(PLDHashTable *table, const PLDHashEntryHdr *entry,
   const AtomTableEntry *he = static_cast<const AtomTableEntry*>(entry);
   const AtomTableEntry *strKey = static_cast<const AtomTableEntry*>(key);
 
-  const char *atomString = he->getAtomString();
+  const PRUnichar *atomString = he->getAtomString();
 
-  if (strKey->IsUTF16String()) {
+  if (strKey->IsUTF8String()) {
     return
-      CompareUTF8toUTF16(nsDependentCSubstring(atomString, atomString + he->getLength()),
+      CompareUTF8toUTF16(nsDependentCSubstring(strKey->getUTF8String(),
-                         nsDependentSubstring(strKey->getUTF16String(),
+                                               strKey->getUTF8String() + strKey->getLength()),
-                                              strKey->getUTF16String() + strKey->getLength())) == 0;
+                         nsDependentSubstring(atomString, atomString + he->getLength())) == 0;
   }
 
   PRUint32 length = he->getLength();
@@ -231,15 +231,15 @@ AtomTableMatchKey(PLDHashTable *table, const PLDHashEntryHdr *entry,
     return PR_FALSE;
   }
 
-  const char *str;
+  const PRUnichar *str;
 
-  if (strKey->IsUTF8String()) {
+  if (strKey->IsUTF16String()) {
-    str = strKey->getUTF8String();
+    str = strKey->getUTF16String();
   } else {
     str = strKey->getAtomString();
   }
 
-  return memcmp(atomString, str, length * sizeof(char)) == 0;
+  return memcmp(atomString, str, length * sizeof(PRUnichar)) == 0;
 }
 
 static void
@@ -298,9 +298,9 @@ DumpAtomLeaks(PLDHashTable *table, PLDHashEntryHdr *he,
   AtomImpl* atom = entry->GetAtomImpl();
   if (!atom->IsPermanent()) {
     ++*static_cast<PRUint32*>(arg);
-    const char *str;
+    nsCAutoString str;
-    atom->GetUTF8String(&str);
+    atom->ToUTF8String(str);
-    fputs(str, stdout);
+    fputs(str.get(), stdout);
     fputs("\n", stdout);
   }
   return PL_DHASH_NEXT;
@@ -343,25 +343,25 @@ NS_PurgeAtomTable()
   }
 }
 
-AtomImpl::AtomImpl(const nsACString& aString)
+AtomImpl::AtomImpl(const nsAString& aString)
   : mLength(aString.Length())
 {
   nsStringBuffer* buf = nsStringBuffer::FromString(aString);
   if (buf) {
     buf->AddRef();
-    mString = static_cast<char*>(buf->Data());
+    mString = static_cast<PRUnichar*>(buf->Data());
   }
   else {
-    buf = nsStringBuffer::Alloc((mLength + 1) * sizeof(char));
+    buf = nsStringBuffer::Alloc((mLength + 1) * sizeof(PRUnichar));
-    mString = static_cast<char*>(buf->Data());
+    mString = static_cast<PRUnichar*>(buf->Data());
-    memcpy(mString, aString.BeginReading(), mLength * sizeof(char));
+    CopyUnicodeTo(aString, 0, mString, mLength);
     mString[mLength] = PRUnichar(0);
   }
 }
 
 AtomImpl::AtomImpl(nsStringBuffer* aStringBuffer, PRUint32 aLength)
   : mLength(aLength),
-    mString(static_cast<char*>(aStringBuffer->Data()))
+    mString(static_cast<PRUnichar*>(aStringBuffer->Data()))
 {
   // Technically we could currently avoid doing this addref by instead making
   // the static atom buffers have an initial refcount of 2.
@@ -430,37 +430,43 @@ void* PermanentAtomImpl::operator new ( size_t size, AtomImpl* aAtom ) CPP_THROW
 NS_IMETHODIMP 
 AtomImpl::ToString(nsAString& aBuf)
 {
-  CopyUTF8toUTF16(nsDependentCString(mString, mLength), aBuf);
+  nsStringBuffer::FromData(mString)->ToString(mLength, aBuf);
   return NS_OK;
 }
 
 NS_IMETHODIMP
 AtomImpl::ToUTF8String(nsACString& aBuf)
 {
-  nsStringBuffer::FromData(mString)->ToString(mLength, aBuf);
+  CopyUTF16toUTF8(nsDependentString(mString, mLength), aBuf);
   return NS_OK;
 }
 
 NS_IMETHODIMP 
-AtomImpl::GetUTF8String(const char **aResult)
+AtomImpl::GetUTF16String(const PRUnichar **aResult)
 {
   NS_PRECONDITION(aResult, "null out param");
   *aResult = mString;
   return NS_OK;
 }
 
+NS_IMETHODIMP_(PRUint32)
+AtomImpl::GetLength()
+{
+  return mLength;
+}
+
 NS_IMETHODIMP
 AtomImpl::EqualsUTF8(const nsACString& aString, PRBool* aResult)
 {
-  *aResult = aString.Equals(nsDependentCString(mString, mLength));
+  *aResult = CompareUTF8toUTF16(aString,
+                                nsDependentString(mString, mLength)) == 0;
   return NS_OK;
 }
 
 NS_IMETHODIMP
 AtomImpl::Equals(const nsAString& aString, PRBool* aResult)
 {
-  *aResult = CompareUTF8toUTF16(nsDependentCString(mString, mLength),
+  *aResult = aString.Equals(nsDependentString(mString, mLength));
-                                aString) == 0;
   return NS_OK;
 }
 
@@ -540,14 +546,15 @@ NS_RegisterStaticAtoms(const nsStaticAtom* aAtoms, PRUint32 aAtomCount)
   }
   
   for (PRUint32 i=0; i<aAtomCount; i++) {
-    NS_ASSERTION(nsCRT::IsAscii((char*)aAtoms[i].mStringBuffer->Data()),
+#ifdef NS_STATIC_ATOM_USE_WIDE_STRINGS
+    NS_ASSERTION(nsCRT::IsAscii((PRUnichar*)aAtoms[i].mStringBuffer->Data()),
                  "Static atoms must be ASCII!");
 
     PRUint32 stringLen =
-      aAtoms[i].mStringBuffer->StorageSize() / sizeof(char) - 1;
+      aAtoms[i].mStringBuffer->StorageSize() / sizeof(PRUnichar) - 1;
 
     AtomTableEntry *he =
-      GetAtomHashEntry((char*)aAtoms[i].mStringBuffer->Data(),
+      GetAtomHashEntry((PRUnichar*)aAtoms[i].mStringBuffer->Data(),
                        stringLen);
 
     if (he->HasValue()) {
@@ -569,11 +576,25 @@ NS_RegisterStaticAtoms(const nsStaticAtom* aAtoms, PRUint32 aAtomCount)
       *aAtoms[i].mAtom = atom;
 
       if (!gStaticAtomTableSealed) {
-        nsAutoString key;
+        gStaticAtomTable->Put(nsAtomString(atom), atom);
-        atom->ToString(key);
-        gStaticAtomTable->Put(key, atom);
       }
     }
+#else // NS_STATIC_ATOM_USE_WIDE_STRINGS
+    NS_ASSERTION(nsCRT::IsAscii((char*)aAtoms[i].mStringBuffer->Data()),
+                 "Static atoms must be ASCII!");
+
+    PRUint32 stringLen = aAtoms[i].mStringBuffer->StorageSize() - 1;
+
+    NS_ConvertASCIItoUTF16 str((char*)aAtoms[i].mStringBuffer->Data(),
+                               stringLen);
+    nsIAtom* atom = NS_NewPermanentAtom(str);
+    *aAtoms[i].mAtom = atom;
+
+    if (!gStaticAtomTableSealed) {
+      gStaticAtomTable->Put(str, atom);
+    }
+#endif
+
   }
   return NS_OK;
 }
@@ -594,13 +615,15 @@ NS_NewAtom(const nsACString& aUTF8String)
     return nsnull;
   }
 
-  NS_ASSERTION(!he->IsUTF8String() && !he->IsUTF16String(),
-               "Atom hash entry is string?  Should be atom!");
-
   if (he->HasValue())
     return he->GetAtom();
 
-  AtomImpl* atom = new AtomImpl(aUTF8String);
+  // This results in an extra addref/release of the nsStringBuffer.
+  // Unfortunately there doesn't seem to be any APIs to avoid that.
+  nsString str;
+  CopyUTF8toUTF16(aUTF8String, str);
+  AtomImpl* atom = new AtomImpl(str);
+
   he->SetAtomImpl(atom);
   NS_ADDREF(atom);
 
@@ -622,12 +645,7 @@ NS_NewAtom(const nsAString& aUTF16String)
   if (he->HasValue())
     return he->GetAtom();
 
-  // This results in an extra addref/release of the nsStringBuffer.
+  AtomImpl* atom = new AtomImpl(aUTF16String);
-  // Unfortunately there doesn't seem to be any APIs to avoid that.
-  nsCString str;
-  CopyUTF16toUTF8(aUTF16String, str);
-  AtomImpl* atom = new AtomImpl(str);
-
   he->SetAtomImpl(atom);
   NS_ADDREF(atom);
 
@@ -635,10 +653,10 @@ NS_NewAtom(const nsAString& aUTF16String)
 }
 
 NS_COM nsIAtom*
-NS_NewPermanentAtom(const nsACString& aUTF8String)
+NS_NewPermanentAtom(const nsAString& aUTF16String)
 {
-  AtomTableEntry *he = GetAtomHashEntry(aUTF8String.Data(),
+  AtomTableEntry *he = GetAtomHashEntry(aUTF16String.Data(),
-                                        aUTF8String.Length());
+                                        aUTF16String.Length());
 
   if (he->HasValue()) {
     AtomImpl* atom = he->GetAtomImpl();
@@ -648,7 +666,7 @@ NS_NewPermanentAtom(const nsACString& aUTF8String)
     return atom;
   }
   
-  AtomImpl* atom = new PermanentAtomImpl(aUTF8String);
+  AtomImpl* atom = new PermanentAtomImpl(aUTF16String);
   he->SetAtomImpl(atom);
 
   // No need to addref since permanent atoms aren't refcounted anyway

xpcom/ds/nsAtomTable.h

@@ -48,7 +48,7 @@
 
 class AtomImpl : public nsIAtom {
 public:
-  AtomImpl(const nsACString& aString);
+  AtomImpl(const nsAString& aString);
 
   // This is currently only used during startup when creating a permanent atom
   // from NS_RegisterStaticAtoms
@@ -60,7 +60,7 @@ protected:
   AtomImpl() {
     // We can't really assert that mString is a valid nsStringBuffer string,
     // so do the best we can do and check for some consistencies.
-    NS_ASSERTION((mLength + 1) * sizeof(char) <=
+    NS_ASSERTION((mLength + 1) * sizeof(PRUnichar) <=
                  nsStringBuffer::FromData(mString)->StorageSize() &&
                  mString[mLength] == 0,
                  "Not initialized atom");
@@ -90,7 +90,7 @@ public:
   PRUint32 mLength;
 
   // This always points to the data owned by a nsStringBuffer
-  char* mString;
+  PRUnichar* mString;
 };
 
 /**
@@ -99,7 +99,7 @@ public:
 
 class PermanentAtomImpl : public AtomImpl {
 public:
-  PermanentAtomImpl(const nsACString& aString)
+  PermanentAtomImpl(const nsAString& aString)
     : AtomImpl(aString)
   {}
   PermanentAtomImpl(nsStringBuffer* aData, PRUint32 aLength)

xpcom/ds/nsCRT.cpp

@@ -55,6 +55,7 @@
 #include "nsIServiceManager.h"
 #include "nsCharTraits.h"
 #include "prbit.h"
+#include "nsUTF8Utils.h"
 
 #define ADD_TO_HASHVAL(hashval, c) \
     hashval = PR_ROTATE_LEFT32(hashval, 4) ^ (c);
@@ -237,125 +238,44 @@ PRUint32 nsCRT::HashCode(const PRUnichar* str, PRUint32* resultingStrLen)
   return h;
 }
 
-PRUint32 nsCRT::HashCodeAsUTF8(const PRUnichar* start, PRUint32 length)
+PRUint32 nsCRT::HashCode(const PRUnichar* start, PRUint32 length)
 {
   PRUint32 h = 0;
   const PRUnichar* s = start;
   const PRUnichar* end = start + length;
 
-  PRUint16 W1 = 0;      // the first UTF-16 word in a two word tuple
+  PRUnichar c;
-  PRUint32 U = 0;       // the current char as UCS-4
+  while ( s < end ) {
-  int code_length = 0;  // the number of bytes in the UTF-8 sequence for the current char
+    c = *s++;
-
+    ADD_TO_HASHVAL(h, c);
-  PRUint16 W;
-  while ( s < end )
-    {
-      W = *s++;
-        /*
-         * On the fly, decoding from UTF-16 (and/or UCS-2) into UTF-8 as per
-         *  http://www.ietf.org/rfc/rfc2781.txt
-         *  http://www.ietf.org/rfc/rfc3629.txt
-         */
-
-      if ( !W1 )
-        {
-          if ( !IS_SURROGATE(W) )
-            {
-              U = W;
-              if ( W <= 0x007F )
-                code_length = 1;
-              else if ( W <= 0x07FF )
-                code_length = 2;
-              else
-                code_length = 3;
-            }
-          else if ( NS_IS_HIGH_SURROGATE(W) && s < end)
-            {
-              W1 = W;
-
-              continue;
-            }
-          else
-            {
-              // Treat broken characters as the Unicode replacement
-              // character 0xFFFD
-              U = 0xFFFD;
-
-              code_length = 3;
-
-              NS_WARNING("Got low surrogate but no previous high surrogate");
-            }
-        }
-      else
-        {
-          // as required by the standard, this code is careful to
-          // throw out illegal sequences
-
-          if ( NS_IS_LOW_SURROGATE(W) )
-            {
-              U = SURROGATE_TO_UCS4(W1, W);
-              NS_ASSERTION(IS_VALID_CHAR(U), "How did this happen?");
-              code_length = 4;
-            }
-          else
-            {
-              // Treat broken characters as the Unicode replacement
-              // character 0xFFFD
-              U = 0xFFFD;
-
-              code_length = 3;
-
-              NS_WARNING("High surrogate not followed by low surrogate");
-
-              // The pointer to the next character points to the second 16-bit
-              // value, not beyond it, as per Unicode 5.0.0 Chapter 3 C10, only
-              // the first code unit of an illegal sequence must be treated as
-              // an illegally terminated code unit sequence (also Chapter 3
-              // D91, "isolated [not paired and ill-formed] UTF-16 code units
-              // in the range D800..DFFF are ill-formed").
-              --s;
-            }
-
-          W1 = 0;
-        }
-
-
-      static const PRUint16 sBytePrefix[5]  = { 0x0000, 0x0000, 0x00C0, 0x00E0, 0x00F0  };
-      static const PRUint16 sShift[5]       = { 0, 0, 6, 12, 18 };
-
-      /*
-       *  Unlike the algorithm in
-       *  http://www.ietf.org/rfc/rfc3629.txt we must calculate the
-       *  bytes in left to right order so that our hash result
-       *  matches what the narrow version would calculate on an
-       *  already UTF-8 string.
-       */
-
-      // hash the first (and often, only, byte)
-      ADD_TO_HASHVAL(h, (sBytePrefix[code_length] | (U>>sShift[code_length])));
-
-      // an unrolled loop for hashing any remaining bytes in this
-      // sequence
-      switch ( code_length )
-        {  // falling through in each case
-          case 4:   ADD_TO_HASHVAL(h, (0x80 | ((U>>12) & 0x003F)));
-          case 3:   ADD_TO_HASHVAL(h, (0x80 | ((U>>6 ) & 0x003F)));
-          case 2:   ADD_TO_HASHVAL(h, (0x80 | ( U      & 0x003F)));
-          default:  code_length = 0;
-            break;
-        }
   }
 
   return h;
 }
 
-PRUint32 nsCRT::BufferHashCode(const PRUnichar* s, PRUint32 len)
+PRUint32 nsCRT::HashCodeAsUTF16(const char* start, PRUint32 length)
 {
   PRUint32 h = 0;
-  const PRUnichar* done = s + len;
+  const char* s = start;
+  const char* end = start + length;
+
+  while ( s < end )
+    {
+      PRBool err;
+      PRUint32 ucs4 = UTF8CharEnumerator::NextChar(&s, end, &err);
+      if (err) {
+	return 0;
+      }
+
+      if (ucs4 < PLANE1_BASE) {
+        ADD_TO_HASHVAL(h, ucs4);
+      }
+      else {
+        ADD_TO_HASHVAL(h, H_SURROGATE(ucs4));
+        ADD_TO_HASHVAL(h, L_SURROGATE(ucs4));
+      }
+    }
 
-  while ( s < done )
-    h = PR_ROTATE_LEFT32(h, 4) ^ PRUint16(*s++); // cast to unsigned to prevent possible sign extension
   return h;
 }
 

xpcom/ds/nsCRT.h

@@ -236,15 +236,13 @@ public:
   static PRUint32 HashCode(const PRUnichar* str,
                            PRUint32* resultingStrLen = nsnull);
 
-  // Computes a hashcode for a length number of UTF16
-  // characters. Returns the same hash code as the HashCode method
-  // taking a |char*| would if the string were converted to UTF8. This
-  // hash function treats invalid UTF16 data as 0xFFFD (0xEFBFBD in
-  // UTF-8).
-  static PRUint32 HashCodeAsUTF8(const PRUnichar* start, PRUint32 length);
-
   // Computes the hashcode for a buffer with a specified length.
-  static PRUint32 BufferHashCode(const PRUnichar* str, PRUint32 strLen);
+  static PRUint32 HashCode(const PRUnichar* str, PRUint32 strLen);
+
+  // Computes a hashcode for a length number of UTF8
+  // characters. Returns the same hash code as the HashCode method
+  // taking a |PRUnichar*| would if the string were converted to UTF16.
+  static PRUint32 HashCodeAsUTF16(const char* start, PRUint32 length);
 
   // String to longlong
   static PRInt64 atoll(const char *str);

xpcom/ds/nsIAtom.idl

@@ -48,7 +48,7 @@
  * pointer identity.
  */
  
-[scriptable, uuid(fbffe332-0856-421a-9b83-aaed0081fc40)]
+[scriptable, uuid(96c82146-56f3-4b43-817f-25d6db1ad8e8)]
 interface nsIAtom : nsISupports
 {
   /**
@@ -58,9 +58,10 @@ interface nsIAtom : nsISupports
   AUTF8String toUTF8String();
   
   /**
-   * Return a pointer to a zero terminated UTF8 string.
+   * Return a pointer to a zero terminated UTF16 string.
    */
-  [noscript] void getUTF8String([shared, retval] out string aResult);
+  [noscript] void getUTF16String([shared, retval] out wstring aResult);
+  [notxpcom] unsigned long getLength();
 
   /**
    * Compare the atom to a specific string value
@@ -83,6 +84,13 @@ interface nsIAtom : nsISupports
     EqualsUTF8(s, &result);
     return result;
   }
+
+  inline const PRUnichar* GetUTF16String() {
+    const PRUnichar* result;
+    GetUTF16String(&result);
+    return result;
+  }
+
 %}
   
   /**
@@ -123,7 +131,6 @@ inline already_AddRefed<nsIAtom> do_GetAtom(const char* aUTF8String)
  * Find an atom that matches the given UTF-8 string.
  */
 extern NS_COM nsIAtom* NS_NewAtom(const nsACString& aUTF8String);
-extern NS_COM nsIAtom* NS_NewPermanentAtom(const nsACString& aUTF8String);
 inline already_AddRefed<nsIAtom> do_GetAtom(const nsACString& aUTF8String)
     { return NS_NewAtom(aUTF8String); }
 
@@ -139,6 +146,7 @@ inline already_AddRefed<nsIAtom> do_GetAtom(const PRUnichar* aUTF16String)
  * Find an atom that matches the given UTF-16 string.
  */
 extern NS_COM nsIAtom* NS_NewAtom(const nsAString& aUTF16String);
+extern NS_COM nsIAtom* NS_NewPermanentAtom(const nsAString& aUTF16String);
 inline already_AddRefed<nsIAtom> do_GetAtom(const nsAString& aUTF16String)
     { return NS_NewAtom(aUTF16String); }
 
@@ -159,4 +167,31 @@ extern NS_COM nsIAtom* NS_GetStaticAtom(const nsAString& aUTF16String);
  */
 extern NS_COM void NS_SealStaticAtomTable();
 
+class nsAtomString : public nsString
+{
+public:
+  nsAtomString(nsIAtom* aAtom)
+  {
+    aAtom->ToString(*this);
+  }
+};
+
+class nsAtomCString : public nsCString
+{
+public:
+  nsAtomCString(nsIAtom* aAtom)
+  {
+    aAtom->ToUTF8String(*this);
+  }
+};
+
+class nsDependentAtomString : public nsDependentString
+{
+public:
+  nsDependentAtomString(nsIAtom* aAtom)
+    : nsDependentString(aAtom->GetUTF16String(), aAtom->GetLength())
+  {
+  }
+};
+
 %}

xpcom/ds/nsStaticAtom.h

@@ -43,8 +43,18 @@
 #include "nsStringBuffer.h"
 #include "prlog.h"
 
+#if defined(HAVE_CPP_CHAR16_T)
+#define NS_STATIC_ATOM_USE_WIDE_STRINGS
+typedef char16_t nsStaticAtomStringType;
+#elif defined(HAVE_CPP_2BYTE_WCHAR_T)
+#define NS_STATIC_ATOM_USE_WIDE_STRINGS
+typedef wchar_t nsStaticAtomStringType;
+#else
+typedef char nsStaticAtomStringType;
+#endif
+
 #define NS_STATIC_ATOM(buffer_name, atom_ptr)  { (nsStringBuffer*) &buffer_name, atom_ptr }
-#define NS_STATIC_ATOM_BUFFER(buffer_name, str_data) static nsFakeStringBuffer< sizeof(str_data) > buffer_name = { 1, sizeof(str_data), str_data };
+#define NS_STATIC_ATOM_BUFFER(buffer_name, str_data) static nsFakeStringBuffer< sizeof(str_data) > buffer_name = { 1, sizeof(str_data) * sizeof(nsStaticAtomStringType), NS_L(str_data) };
 
 /**
  * Holds data used to initialize large number of atoms during startup. Use
@@ -63,11 +73,10 @@ template <PRUint32 size>
 struct nsFakeStringBuffer {
     PRInt32 mRefCnt;
     PRUint32 mSize;
-    char mStringData[size];
+    nsStaticAtomStringType mStringData[size];
 };
 
-// register your lookup function with the atom table. Your function
+// Register static atoms with the atom table
-// will be called when at atom is not found in the main atom table.
 NS_COM nsresult
 NS_RegisterStaticAtoms(const nsStaticAtom*, PRUint32 aAtomCount);
 

xpcom/tests/TestAtoms.cpp

@@ -81,9 +81,9 @@ int main(int argc, char** argv)
 
   // Now make sure we can find all the idents we just made
   for (i = 0; i < count; i++) {
-    const char *utf8String;
+    const PRUnichar *utf16String;
-    ids[i]->GetUTF8String(&utf8String);
+    ids[i]->GetUTF16String(&utf16String);
-    nsIAtom* id = NS_NewAtom(utf8String);
+    nsIAtom* id = NS_NewAtom(utf16String);
     if (id != ids[i]) {
       id->ToString(s1);
       ids[i]->ToString(s2);

xpcom/tests/TestPermanentAtoms.cpp

@@ -51,9 +51,7 @@ static void Assert(PRBool aCondition, const char* aStatement)
 
 static void AssertString(nsIAtom *aAtom, const nsACString& aString)
 {
-    const char *str;
+    Assert(aAtom->EqualsUTF8(aString), "string is correct");
-    static_cast<AtomImpl*>(aAtom)->GetUTF8String(&str);
-    Assert(nsDependentCString(str) == aString, "string is correct");
 }
 
 static void AssertPermanence(nsIAtom *aAtom, PRBool aPermanence)
@@ -68,13 +66,13 @@ int main()
     AssertString(foo, NS_LITERAL_CSTRING("foo"));
     AssertPermanence(foo, PR_FALSE);
 
-    nsCOMPtr<nsIAtom> foop = do_GetPermanentAtom("foo");
+    nsCOMPtr<nsIAtom> foop = NS_NewPermanentAtom(NS_LITERAL_STRING("foo"));
     AssertString(foop, NS_LITERAL_CSTRING("foo"));
     AssertPermanence(foop, PR_TRUE);
     
     Assert(foo == foop, "atoms are equal");
     
-    nsCOMPtr<nsIAtom> barp = do_GetPermanentAtom("bar");
+    nsCOMPtr<nsIAtom> barp = NS_NewPermanentAtom(NS_LITERAL_STRING("bar"));
     AssertString(barp, NS_LITERAL_CSTRING("bar"));
     AssertPermanence(barp, PR_TRUE);
     

xpcom/tests/TestUTF.cpp

@@ -140,6 +140,32 @@ test_malformed8()
   return PR_TRUE;
 }
 
+PRBool
+test_hashas16()
+{
+  for (unsigned int i = 0; i < NS_ARRAY_LENGTH(ValidStrings); ++i) {
+    nsDependentCString str8(ValidStrings[i].m8);
+    if (nsCRT::HashCode(ValidStrings[i].m16) !=
+        nsCRT::HashCodeAsUTF16(str8.get(), str8.Length()))
+      return PR_FALSE;
+  }
+
+  for (unsigned int i = 0; i < NS_ARRAY_LENGTH(Invalid8Strings); ++i) {
+    nsDependentCString str8(Invalid8Strings[i].m8);
+    if (nsCRT::HashCode(Invalid8Strings[i].m16) !=
+        nsCRT::HashCodeAsUTF16(str8.get(), str8.Length()))
+      return PR_FALSE;
+  }
+
+  for (unsigned int i = 0; i < NS_ARRAY_LENGTH(Malformed8Strings); ++i) {
+    nsDependentCString str8(Malformed8Strings[i]);
+    if (nsCRT::HashCodeAsUTF16(str8.get(), str8.Length()) != 0)
+      return PR_FALSE;
+  }
+
+  return PR_TRUE;
+}
+
 typedef PRBool (*TestFunc)();
 
 static const struct Test
@@ -156,6 +182,7 @@ tests[] =
     // Don't run this test in debug builds as that intentionally asserts
     { "test_malformed8", test_malformed8 },
 #endif
+    { "test_hashas16", test_hashas16 },
     { nsnull, nsnull }
   };