fix the problem with parsing HTML comments (closes bug 1116708; based on patch 1168583)

git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/trunk@45336 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775

include/wx/html/htmlpars.h

@@ -128,6 +128,13 @@ public:
     // Returns entity parser object, used to substitute HTML &entities;
     wxHtmlEntitiesParser *GetEntitiesParser() const { return m_entitiesParser; }
 
+    // Returns true if the tag starting at the given position is a comment tag
+    //
+    // p should point to '<' character and is modified to point to the closing
+    // '>' of the end comment tag if this is indeed a comment
+    static bool
+    SkipCommentTag(wxString::const_iterator& p, wxString::const_iterator end);
+
 protected:
     // DOM structure
     void CreateDOMTree();

src/html/htmlpars.cpp

@@ -171,29 +171,11 @@ void wxHtmlParser::CreateDOMSubTree(wxHtmlTag *cur,
                     wxHtmlTextPiece(textBeginning, i - textBeginning));
 
             // if it is a comment, skip it:
-            if (i < end_pos-6 && m_Source.GetChar(i+1) == wxT('!') &&
+            wxString::const_iterator iter = m_Source.begin() + i;
-                                 m_Source.GetChar(i+2) == wxT('-') &&
+            if ( SkipCommentTag(iter, m_Source.end()) )
-                                 m_Source.GetChar(i+3) == wxT('-'))
             {
-                // Comments begin with "<!--" and end with "--[ \t\r\n]*>"
+                textBeginning =
-                // according to HTML 4.0
+                i = iter - m_Source.begin() + 1; // skip closing '>' too
-                int dashes = 0;
-                i += 4;
-                while (i < end_pos)
-                {
-                    c = m_Source.GetChar(i++);
-                    if ((c == wxT(' ') || c == wxT('\n') ||
-                        c == wxT('\r') || c == wxT('\t')) && dashes >= 2) {}
-                    else if (c == wxT('>') && dashes >= 2)
-                    {
-                        textBeginning = i;
-                        break;
-                    }
-                    else if (c == wxT('-'))
-                        dashes++;
-                    else
-                        dashes = 0;
-                }
             }
 
             // add another tag to the tree:
@@ -951,4 +933,55 @@ wxString wxHtmlParser::ExtractCharsetInformation(const wxString& markup)
     return charset;
 }
 
-#endif
+/* static */
+bool
+wxHtmlParser::SkipCommentTag(wxString::const_iterator& start,
+                             wxString::const_iterator end)
+{
+    wxASSERT_MSG( *start == '<', _T("should be called on the tag start") );
+
+    wxString::const_iterator p = start;
+
+    // comments begin with "<!--" in HTML 4.0
+    if ( end - p < 3 || *++p != '!' || *++p != '-' || *++p != '-' )
+    {
+        // not a comment at all
+        return false;
+    }
+
+    // skip the start of the comment tag in any case, if we don't find the
+    // closing tag we should ignore broken markup
+    start = p;
+
+    // comments end with "--[ \t\r\n]*>", i.e. white space is allowed between
+    // comment delimiter and the closing tag character (section 3.2.4 of
+    // http://www.w3.org/TR/html401/)
+    int dashes = 0;
+    while ( ++p < end )
+    {
+        const wxChar c = *p;
+
+        if ( (c == wxT(' ') || c == wxT('\n') ||
+              c == wxT('\r') || c == wxT('\t')) && dashes >= 2 )
+        {
+            // ignore white space before potential tag end
+            continue;
+        }
+
+        if ( c == wxT('>') && dashes >= 2 )
+        {
+            // found end of comment
+            start = p;
+            break;
+        }
+
+        if ( c == wxT('-') )
+            dashes++;
+        else
+            dashes = 0;
+    }
+
+    return true;
+}
+
+#endif // wxUSE_HTML

src/html/htmltag.cpp

@@ -68,11 +68,18 @@ wxHtmlTagsCache::wxHtmlTagsCache(const wxString& source)
     m_CacheSize = 0;
     m_CachePos = 0;
 
-    int pos = 0;
+    for ( int pos = 0; pos < lng; pos++ )
-    while (pos < lng)
     {
         if (src[pos] == wxT('<'))   // tag found:
         {
+            // don't cache comment tags
+            wxString::const_iterator iter = source.begin() + pos;
+            if ( wxHtmlParser::SkipCommentTag(iter, source.end()) )
+            {
+                pos = iter - source.begin();
+                continue;
+            }
+
             if (m_CacheSize % CACHE_INCREMENT == 0)
                 m_Cache = (wxHtmlCacheItem*) realloc(m_Cache, (m_CacheSize + CACHE_INCREMENT) * sizeof(wxHtmlCacheItem));
             int tg = m_CacheSize++;
@@ -169,8 +176,6 @@ wxHtmlTagsCache::wxHtmlTagsCache(const wxString& source)
                 }
             }
         }
-
-        pos++;
     }
 
     // ok, we're done, now we'll free .Name members of cache - we don't need it anymore: