diff --git a/Lib/markupbase.py b/Lib/markupbase.py
index 24808d185be..ddeb9835b80 100644
--- a/Lib/markupbase.py
+++ b/Lib/markupbase.py
@@ -108,6 +108,10 @@ class ParserBase:
if decltype == "doctype":
self.handle_decl(data)
else:
+ # According to the HTML5 specs sections "8.2.4.44 Bogus
+ # comment state" and "8.2.4.45 Markup declaration open
+ # state", a comment token should be emitted.
+ # Calling unknown_decl provides more flexibility though.
self.unknown_decl(data)
return j + 1
if c in "\"'":
diff --git a/Lib/test/test_htmlparser.py b/Lib/test/test_htmlparser.py
index 5dfe466225e..14ed80c5d09 100644
--- a/Lib/test/test_htmlparser.py
+++ b/Lib/test/test_htmlparser.py
@@ -449,6 +449,48 @@ class AttributesTestCase(TestCaseBase):
[("href", "http://www.example.org/\">;")]),
("data", "spam"), ("endtag", "a")])
+ def test_condcoms(self):
+ html = (''
+ ''
+ '')
+ expected = [('comment', "[if IE & !(lte IE 8)]>aren'tcondcomspretty?'
+ html = ('broken condcom'
+ ''
+ '
'
+ 'foo'
+ '
')
+ # According to the HTML5 specs sections "8.2.4.44 Bogus comment state"
+ # and "8.2.4.45 Markup declaration open state", comment tokens should
+ # be emitted instead of 'unknown decl', but calling unknown_decl
+ # provides more flexibility.
+ # See also Lib/_markupbase.py:parse_declaration
+ expected = [
+ ('unknown decl', 'if !(IE)'),
+ ('data', 'broken condcom'),
+ ('unknown decl', 'endif'),
+ ('unknown decl', 'if ! IE'),
+ ('startendtag', 'link', [('href', 'favicon.tiff')]),
+ ('unknown decl', 'endif'),
+ ('unknown decl', 'if !IE 6'),
+ ('startendtag', 'img', [('src', 'firefox.png')]),
+ ('unknown decl', 'endif'),
+ ('unknown decl', 'if !ie 6'),
+ ('starttag', 'b', []),
+ ('data', 'foo'),
+ ('endtag', 'b'),
+ ('unknown decl', 'endif'),
+ ('unknown decl', 'if (!IE)|(lt IE 9)'),
+ ('startendtag', 'img', [('src', 'mammoth.bmp')]),
+ ('unknown decl', 'endif')
+ ]
+ self._run_check(html, expected)
+
def test_main():
test_support.run_unittest(HTMLParserTestCase, AttributesTestCase)