fix(dep): HTML parsing of processing instructions

Added test coverage to describe behavior of our parsers, and update nekohtml to 1.9.22.noko2.
sparklemotion · Apr 10, 2022 · 0feac5a · 0feac5a
1 parent db72b90
commit 0feac5a
Show file tree

Hide file tree

Showing 3 changed files with 35 additions and 8 deletions.
diff --git a/lib/nekohtml.jar b/lib/nekohtml.jar
diff --git a/test/html4/test_document.rb b/test/html4/test_document.rb
@@ -727,6 +727,17 @@ def test_silencing_nonparse_errors_during_attribute_insertion_1262
           assert_equal(0, doc.errors.length)
         end
 
+        def test_leaking_dtd_nodes_after_internal_subset_removal
+          # see https://github.com/sparklemotion/nokogiri/issues/1784
+          #
+          # just checking that this doesn't raise a valgrind error. we
+          # don't otherwise have any test coverage for removing DTDs.
+          #
+          100.times do |_i|
+            Nokogiri::HTML::Document.new.internal_subset.remove
+          end
+        end
+
         it "skips encoding for script tags" do
           html = Nokogiri::HTML(<<~EOHTML)
             <html>
@@ -763,15 +774,17 @@ def test_silencing_nonparse_errors_during_attribute_insertion_1262
           assert_equal "ISO-8859-1", html.encoding.name
         end
 
-        def test_leaking_dtd_nodes_after_internal_subset_removal
-          # see https://github.com/sparklemotion/nokogiri/issues/1784
-          #
-          # just checking that this doesn't raise a valgrind error. we
-          # don't otherwise have any test coverage for removing DTDs.
-          #
-          100.times do |_i|
-            Nokogiri::HTML::Document.new.internal_subset.remove
+        it "handles ill-formed processing instructions" do
+          html = %{<html><body><!--><?a/}
+          doc = Nokogiri::HTML4::Document.parse(html)
+          expected = if Nokogiri.jruby?
+            [Nokogiri::XML::Node::COMMENT_NODE, Nokogiri::XML::Node::PI_NODE]
+          elsif Nokogiri.libxml2_patches.include?("0008-htmlParseComment-handle-abruptly-closed-comments.patch")
+            [Nokogiri::XML::Node::COMMENT_NODE]
+          else
+            []
           end
+          assert_equal(expected, doc.at_css("body").children.map(&:type))
         end
 
         describe ".parse" do

diff --git a/test/html5/test_nokogumbo.rb b/test/html5/test_nokogumbo.rb
@@ -322,6 +322,20 @@ def test_line_cdata
     assert_equal(3, node.line)
   end
 
+  it "handles ill-formed processing instructions in a document" do
+    html = %{<html><body><!--><?a/}
+    doc = Nokogiri::HTML5::Document.parse(html)
+    expected = [Nokogiri::XML::Node::COMMENT_NODE, Nokogiri::XML::Node::COMMENT_NODE]
+    assert_equal(expected, doc.at_css("body").children.map(&:type))
+  end
+
+  it "handles ill-formed processing instructions in a fragment" do
+    html = %{<div><!--><?a/}
+    frag = Nokogiri::HTML5::DocumentFragment.parse(html)
+    expected = [Nokogiri::XML::Node::COMMENT_NODE, Nokogiri::XML::Node::COMMENT_NODE]
+    assert_equal(expected, frag.at_css("div").children.map(&:type))
+  end
+
   private
 
   def buffer