Bug 1749935 - Always create scanner upfront. r=hsivonen

Differential Revision: https://phabricator.services.mozilla.com/D135880

parser/htmlparser/CParserContext.cpp

@@ -12,17 +12,29 @@
 #include "nsMimeTypes.h"
 #include "nsHTMLTokenizer.h"
 
-CParserContext::CParserContext(nsScanner* aScanner, eParserCommands aCommand,
-                               eAutoDetectResult aStatus, bool aCopyUnused)
-    : mScanner(mozilla::WrapUnique(aScanner)),
+CParserContext::CParserContext(nsIURI* aURI, eParserCommands aCommand)
+    : mScanner(aURI),
       mDTDMode(eDTDMode_autodetect),
-      mDocType(static_cast<eParserDocType>(0)),
+      mDocType(eUnknown),
       mStreamListenerState(eNone),
-      mContextType(eCTNone),
-      mAutoDetectStatus(aStatus),
+      mContextType(eCTURL),
       mParserCommand(aCommand),
       mMultipart(true),
-      mCopyUnused(aCopyUnused) {
+      mCopyUnused(false) {
+  MOZ_COUNT_CTOR(CParserContext);
+}
+
+CParserContext::CParserContext(const nsAString& aBuffer,
+                               eParserCommands aCommand, bool aLastBuffer)
+    : mScanner(aBuffer, !aLastBuffer),
+      mMimeType("application/xml"_ns),
+      mDTDMode(eDTDMode_full_standards),
+      mDocType(eXML),
+      mStreamListenerState(aLastBuffer ? eOnStop : eOnDataAvail),
+      mContextType(eCTString),
+      mParserCommand(aCommand),
+      mMultipart(!aLastBuffer),
+      mCopyUnused(aLastBuffer) {
   MOZ_COUNT_CTOR(CParserContext);
 }
 
@@ -34,7 +46,7 @@ CParserContext::~CParserContext() {
 void CParserContext::SetMimeType(const nsACString& aMimeType) {
   mMimeType.Assign(aMimeType);
 
-  mDocType = ePlainText;
+  mDocType = eUnknown;
 
   if (mMimeType.EqualsLiteral(TEXT_HTML))
     mDocType = eHTML_Strict;

parser/htmlparser/CParserContext.h

@@ -29,12 +29,11 @@ class nsITokenizer;
 
 class CParserContext {
  public:
-  enum eContextType { eCTNone, eCTURL, eCTString, eCTStream };
+  enum eContextType { eCTURL, eCTString };
 
-  explicit CParserContext(nsScanner* aScanner,
-                          eParserCommands aCommand = eViewNormal,
-                          eAutoDetectResult aStatus = eUnknownDetect,
-                          bool aCopyUnused = false);
+  CParserContext(nsIURI* aURI, eParserCommands aCommand);
+  CParserContext(const nsAString& aBuffer, eParserCommands aCommand,
+                 bool aLastBuffer);
 
   ~CParserContext();
 
@@ -46,15 +45,15 @@ class CParserContext {
       mRequest;  // provided by necko to differnciate different input streams
                  // why is mRequest strongly referenced? see bug 102376.
   nsCOMPtr<nsITokenizer> mTokenizer;
-  mozilla::UniquePtr<nsScanner> mScanner;
+  nsScanner mScanner;
 
   nsCString mMimeType;
   nsDTDMode mDTDMode;
 
   eParserDocType mDocType;
   eStreamState mStreamListenerState;
   eContextType mContextType;
-  eAutoDetectResult mAutoDetectStatus;
+  eAutoDetectResult mAutoDetectStatus = eUnknownDetect;
   eParserCommands mParserCommand;
 
   bool mMultipart;

parser/htmlparser/nsExpatDriver.cpp

@@ -1576,7 +1576,7 @@ nsExpatDriver::WillBuildModel(const CParserContext& aParserContext,
                     XML_PARAM_ENTITY_PARSING_ALWAYS);
 #endif
 
-  auto baseURI = GetExpatBaseURI(aParserContext.mScanner->GetURI());
+  auto baseURI = GetExpatBaseURI(aParserContext.mScanner.GetURI());
   auto uri =
       TransferBuffer<XML_Char>(Sandbox(), &baseURI[0], ArrayLength(baseURI));
   RLBOX_EXPAT_MCALL(MOZ_XML_SetBase, *uri);

parser/htmlparser/nsIParser.h

@@ -43,7 +43,7 @@ class Encoding;
 
 enum eParserCommands { eViewNormal, eViewSource, eViewFragment, eViewErrors };
 
-enum eParserDocType { ePlainText = 0, eXML, eHTML_Quirks, eHTML_Strict };
+enum eParserDocType { eUnknown = 0, eXML, eHTML_Quirks, eHTML_Strict };
 
 enum eStreamState { eNone, eOnStart, eOnDataAvail, eOnStop };
 

parser/htmlparser/nsParser.cpp

@@ -234,8 +234,8 @@ void nsParser::SetDocumentCharset(NotNull<const Encoding*> aCharset,
                                   bool aForceAutoDetection) {
   mCharset = aCharset;
   mCharsetSource = aCharsetSource;
-  if (mParserContext && mParserContext->mScanner) {
-    mParserContext->mScanner->SetDocumentCharset(aCharset, aCharsetSource);
+  if (mParserContext) {
+    mParserContext->mScanner.SetDocumentCharset(aCharset, aCharsetSource);
   }
 }
 
@@ -547,10 +547,7 @@ nsParser::Parse(nsIURI* aURL) {
 
   MOZ_ASSERT(!mParserContext, "We expect mParserContext to be null.");
 
-  nsScanner* theScanner = new nsScanner(aURL, false);
-  mParserContext = MakeUnique<CParserContext>(theScanner, mCommand);
-  mParserContext->mMultipart = true;
-  mParserContext->mContextType = CParserContext::eCTURL;
+  mParserContext = MakeUnique<CParserContext>(aURL, mCommand);
 
   return NS_OK;
 }
@@ -561,8 +558,6 @@ nsParser::Parse(nsIURI* aURL) {
  * @param   aSourceBuffer contains a string-full of real content
  */
 nsresult nsParser::Parse(const nsAString& aSourceBuffer, bool aLastCall) {
-  nsresult result = NS_OK;
-
   if (mInternalState == NS_ERROR_OUT_OF_MEMORY) {
     // Checking NS_ERROR_OUT_OF_MEMORY instead of NS_FAILED
     // to avoid introducing unintentional changes to behavior.
@@ -571,7 +566,7 @@ nsresult nsParser::Parse(const nsAString& aSourceBuffer, bool aLastCall) {
 
   // Don't bother if we're never going to parse this.
   if (mInternalState == NS_ERROR_HTMLPARSER_STOPPARSING) {
-    return result;
+    return NS_OK;
   }
 
   if (!aLastCall && aSourceBuffer.IsEmpty()) {
@@ -580,61 +575,28 @@ nsresult nsParser::Parse(const nsAString& aSourceBuffer, bool aLastCall) {
     // some data is actually passed in.
     // But if this is the last call, make sure to finish up
     // stuff correctly.
-    return result;
+    return NS_OK;
   }
 
   // Maintain a reference to ourselves so we don't go away
   // till we're completely done.
   nsCOMPtr<nsIParser> kungFuDeathGrip(this);
 
-  if (aLastCall || !aSourceBuffer.IsEmpty() || !mUnusedInput.IsEmpty()) {
-    if (!mParserContext) {
-      // Only make a new context if we don't have one.
-      nsScanner* theScanner = new nsScanner(mUnusedInput);
-      NS_ENSURE_TRUE(theScanner, NS_ERROR_OUT_OF_MEMORY);
-
-      mParserContext = MakeUnique<CParserContext>(theScanner, mCommand,
-                                                  eUnknownDetect, aLastCall);
-
-      mParserContext->mMultipart = !aLastCall;  // By default
-
-      // Start fix bug 40143
-      if (mParserContext->mMultipart) {
-        mParserContext->mStreamListenerState = eOnDataAvail;
-        if (mParserContext->mScanner) {
-          mParserContext->mScanner->SetIncremental(true);
-        }
-      } else {
-        mParserContext->mStreamListenerState = eOnStop;
-        if (mParserContext->mScanner) {
-          mParserContext->mScanner->SetIncremental(false);
-        }
-      }
-      // end fix for 40143
-
-      mParserContext->mContextType = CParserContext::eCTString;
-      mParserContext->SetMimeType("application/xml"_ns);
-      mParserContext->mDTDMode = eDTDMode_full_standards;
-
-      mUnusedInput.Truncate();
-
-      mParserContext->mScanner->Append(aSourceBuffer);
-      // Do not interrupt document.write() - bug 95487
-      result = ResumeParse(false, false, false);
-    } else {
-      mParserContext->mScanner->Append(aSourceBuffer);
-      // Set stream listener state to eOnStop, on the final context - Fix
-      // 68160, to guarantee DidBuildModel() call - Fix 36148
-      if (aLastCall) {
-        mParserContext->mStreamListenerState = eOnStop;
-        mParserContext->mScanner->SetIncremental(false);
-      }
+  if (!mParserContext) {
+    // Only make a new context if we don't have one.
+    mParserContext =
+        MakeUnique<CParserContext>(mUnusedInput, mCommand, aLastCall);
 
-      ResumeParse(false, false, false);
-    }
+    mUnusedInput.Truncate();
+  } else if (aLastCall) {
+    // Set stream listener state to eOnStop, on the final context - Fix
+    // 68160, to guarantee DidBuildModel() call - Fix 36148
+    mParserContext->mStreamListenerState = eOnStop;
+    mParserContext->mScanner.SetIncremental(false);
   }
 
-  return result;
+  mParserContext->mScanner.Append(aSourceBuffer);
+  return ResumeParse(false, false, false);
 }
 
 NS_IMETHODIMP
@@ -765,12 +727,12 @@ nsresult nsParser::ResumeParse(bool allowIteration, bool aIsFinalChunk,
       bool theIterationIsOk = true;
 
       while (result == NS_OK && theIterationIsOk) {
-        if (!mUnusedInput.IsEmpty() && mParserContext->mScanner) {
+        if (!mUnusedInput.IsEmpty()) {
           // -- Ref: Bug# 22485 --
           // Insert the unused input into the source buffer
           // as if it was read from the input stream.
           // Adding UngetReadable() per vidur!!
-          mParserContext->mScanner->UngetReadable(mUnusedInput);
+          mParserContext->mScanner.UngetReadable(mUnusedInput);
           mUnusedInput.Truncate(0);
         }
 
@@ -1089,7 +1051,7 @@ nsresult nsParser::OnDataAvailable(nsIRequest* request,
     ParserWriteStruct pws;
     pws.mNeedCharsetCheck = true;
     pws.mParser = this;
-    pws.mScanner = mParserContext->mScanner.get();
+    pws.mScanner = &mParserContext->mScanner;
     pws.mRequest = request;
 
     rv = pIStream->ReadSegments(ParserWriteFunc, &pws, aLength, &totalRead);
@@ -1129,7 +1091,7 @@ nsresult nsParser::OnStopRequest(nsIRequest* request, nsresult status) {
 
   if (mParserContext->mRequest == request) {
     mParserContext->mStreamListenerState = eOnStop;
-    mParserContext->mScanner->SetIncremental(false);
+    mParserContext->mScanner.SetIncremental(false);
   }
 
   mStreamStatus = status;
@@ -1193,10 +1155,10 @@ nsresult nsParser::Tokenize(bool aIsFinalChunk) {
 
     WillTokenize(aIsFinalChunk);
     while (NS_SUCCEEDED(result)) {
-      mParserContext->mScanner->Mark();
-      result = theTokenizer->ConsumeToken(*mParserContext->mScanner);
+      mParserContext->mScanner.Mark();
+      result = theTokenizer->ConsumeToken(mParserContext->mScanner);
       if (NS_FAILED(result)) {
-        mParserContext->mScanner->RewindToMark();
+        mParserContext->mScanner.RewindToMark();
         if (NS_ERROR_HTMLPARSER_EOF == result) {
           break;
         }

parser/htmlparser/nsScanner.cpp

@@ -47,7 +47,8 @@ nsReadEndCondition::nsReadEndCondition(const char16_t* aTerminateChars)
  *  @param   aMode represents the parser mode (nav, other)
  *  @return
  */
-nsScanner::nsScanner(const nsAString& anHTMLString) {
+nsScanner::nsScanner(const nsAString& anHTMLString, bool aIncremental)
+    : mIncremental(aIncremental) {
   MOZ_COUNT_CTOR(nsScanner);
 
   mSlidingBuffer = nullptr;
@@ -59,20 +60,15 @@ nsScanner::nsScanner(const nsAString& anHTMLString) {
     mEndPosition = mCurrentPosition;
   }
   mMarkPosition = mCurrentPosition;
-  mIncremental = false;
-  mUnicodeDecoder = nullptr;
-  mCharsetSource = kCharsetUninitialized;
 }
 
 /**
  *  Use this constructor if you want i/o to be based on strings
  *  the scanner receives. If you pass a null filename, you
  *  can still provide data to the scanner via append.
  */
-nsScanner::nsScanner(nsIURI* aURI, bool aCreateStream) : mURI(aURI) {
+nsScanner::nsScanner(nsIURI* aURI) : mURI(aURI), mIncremental(true) {
   MOZ_COUNT_CTOR(nsScanner);
-  NS_ASSERTION(!aCreateStream, "This is always true.");
-
   mSlidingBuffer = nullptr;
 
   // XXX This is a big hack.  We need to initialize the iterators to something.
@@ -84,10 +80,6 @@ nsScanner::nsScanner(nsIURI* aURI, bool aCreateStream) : mURI(aURI) {
   mMarkPosition = mCurrentPosition;
   mEndPosition = mCurrentPosition;
 
-  mIncremental = true;
-
-  mUnicodeDecoder = nullptr;
-  mCharsetSource = kCharsetUninitialized;
   // XML defaults to UTF-8 and about:blank is UTF-8, too.
   SetDocumentCharset(UTF_8_ENCODING, kCharsetFromDocTypeDefault);
 }

parser/htmlparser/nsScanner.h

@@ -17,6 +17,7 @@
 #ifndef SCANNER
 #define SCANNER
 
+#include "nsCharsetSource.h"
 #include "nsCOMPtr.h"
 #include "nsString.h"
 #include "nsIParser.h"
@@ -44,13 +45,13 @@ class nsScanner final {
   /**
    *  Use this constructor for the XML fragment parsing case
    */
-  explicit nsScanner(const nsAString& anHTMLString);
+  nsScanner(const nsAString& anHTMLString, bool aIncremental);
 
   /**
    *  Use this constructor if you want i/o to be based on
    *  a file (therefore a stream) or just data you provide via Append().
    */
-  nsScanner(nsIURI* aURI, bool aCreateStream);
+  explicit nsScanner(nsIURI* aURI);
 
   ~nsScanner();
 
@@ -176,7 +177,7 @@ class nsScanner final {
   nsScannerIterator mEndPosition;  // The current end of the scanner buffer
   nsCOMPtr<nsIURI> mURI;
   bool mIncremental;
-  int32_t mCharsetSource;
+  int32_t mCharsetSource = kCharsetUninitialized;
   nsCString mCharset;
   mozilla::UniquePtr<mozilla::Decoder> mUnicodeDecoder;