Fix parsing of HTML that has a DOCTYPE declaring it as XHTML but no xmlns attribute. Also coerce downloaded article titles to unicode

src/calibre/ebooks/oeb/base.py

@@ -764,7 +764,14 @@ def _parse_xhtml(self, data):
             # Convert to Unicode and normalize line endings
             data = self.oeb.decode(data)
             data = self.oeb.html_preprocessor(data)
-            orig_data = data
+
+            # Remove DOCTYPE declaration as it messes up parsing
+            # Inparticular it causes a tostring to insert xmlns
+            # declarations, which messes up the coesrcing logic
+            idx = data.find('<html')
+            if idx > -1:
+                data = data[idx:]
+
             # Try with more & more drastic measures to parse
             def first_pass(data):
                 try:

src/calibre/web/feeds/__init__.py

@@ -25,6 +25,8 @@ def __init__(self, id, title, url, author, summary, published, content):
                 entity_to_unicode, self.title)
         except:
             pass
+        if not isinstance(self.title, unicode):
+            self.title = self.title.decode('utf-8', 'replace')
         self.url = url
         self.author = author
         if author and not isinstance(author, unicode):

src/calibre/web/feeds/news.py

@@ -980,7 +980,7 @@ def article_downloaded(self, request, result):
 
     def error_in_article_download(self, request, traceback):
         self.jobs_done += 1
-        self.log.error(_('Failed to download article: %s from %s\n')%(request.article.title, request.article.url))
+        self.log.error(_(u'Failed to download article: %s from %s\n')%(request.article.title, request.article.url))
         self.log.debug(traceback)
         self.log.debug('\n')
         self.report_progress(float(self.jobs_done)/len(self.jobs), _('Article download failed: %s')%request.article.title)