backport updates and fixes to v1.15.x (#2953)

**What problem is this PR intended to solve?**

Backporting changes from:

- #2927
- #2924
- #2951

.github/ISSUE_TEMPLATE/1-bug-report.md

@@ -32,7 +32,7 @@ Here's an example of how you might structure such a script:
 require 'nokogiri'
 require 'minitest/autorun'
 
-class Test < MiniTest::Spec
+class Test < Minitest::Spec
   describe "Node#css" do
     it "should find a div using chained classes" do
       html = <<~HEREDOC

.github/workflows/downstream.yml

@@ -47,7 +47,7 @@ jobs:
             command: "bundle exec rake spec"
           - url: https://github.com/SAML-Toolkits/ruby-saml
             name: ruby-saml
-            command: "bundle exec rake test"
+            command: "bundle exec rake test MT_COMPAT=t"
           # - url: https://github.com/instructure/nokogiri-xmlsec-instructure
           #   name: nokogiri-xmlsec-instructure
           #   precommand: "apt install -y libxmlsec1-dev"

CHANGELOG.md

@@ -4,6 +4,19 @@ Nokogiri follows [Semantic Versioning](https://semver.org/), please see the [REA
 
 ---
 
+## next / unreleased
+
+### Dependencies
+
+* [CRuby] Vendored libxml2 is updated to v2.11.5 from v2.11.4. For details please see https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.11.5
+
+
+### Fixed
+
+* Fixed a typo in a HTML5 parser error message. [[#2927](https://github.com/sparklemotion/nokogiri/issues/2927)] (Thanks, [@anishathalye](https://github.com/anishathalye)!)
+* [CRuby] `ObjectSpace.memsize_of` is now safe to call on `Document`s with complex DTDs. In previous versions, this debugging method could result in a segfault. [[#2923](https://github.com/sparklemotion/nokogiri/issues/2923), [#2924](https://github.com/sparklemotion/nokogiri/issues/2924)]
+
+
 ## 1.15.3 / 2023-07-05
 
 ### Fixed

CONTRIBUTING.md

@@ -124,7 +124,7 @@ Note that `rake test` does not compile the native extension, and this is intenti
 bundle exec rake compile test
 ```
 
-To run a focused test, use MiniTest's `TESTOPTS`:
+To run a focused test, use Minitest's `TESTOPTS`:
 
 ``` sh
 bundle exec rake compile test TESTOPTS="-n/test_last_element_child/"

dependencies.yml

@@ -1,7 +1,7 @@
 libxml2:
-  version: "2.11.4"
-  sha256: "737e1d7f8ab3f139729ca13a2494fd17bf30ddb4b7a427cf336252cab57f57f7"
-  # sha-256 hash provided in https://download.gnome.org/sources/libxml2/2.11/libxml2-2.11.4.sha256sum
+  version: "2.11.5"
+  sha256: "3727b078c360ec69fa869de14bd6f75d7ee8d36987b071e6928d4720a28df3a6"
+  # sha-256 hash provided in https://download.gnome.org/sources/libxml2/2.11/libxml2-2.11.5.sha256sum
 
 libxslt:
   version: "1.1.38"

ext/nokogiri/xml_document.c

@@ -100,11 +100,15 @@ memsize_node(const xmlNodePtr node)
 {
   /* note we don't count namespace definitions, just going for a good-enough number here */
   xmlNodePtr child;
+  xmlAttrPtr property;
   size_t memsize = 0;
 
   memsize += xmlStrlen(node->name);
-  for (child = (xmlNodePtr)node->properties; child; child = child->next) {
-    memsize += sizeof(xmlAttr) + memsize_node(child);
+
+  if (node->type == XML_ELEMENT_NODE) {
+    for (property = node->properties; property; property = property->next) {
+      memsize += sizeof(xmlAttr) + memsize_node((xmlNodePtr)property);
+    }
   }
   if (node->type == XML_TEXT_NODE) {
     memsize += xmlStrlen(node->content);

gumbo-parser/src/error.c

@@ -357,7 +357,7 @@ static void handle_parser_error (
       print_tag_stack(error, output);
       return;
     case GUMBO_TOKEN_END_TAG:
-      print_message(output, "Eng tag '%s' isn't allowed here.",
+      print_message(output, "End tag '%s' isn't allowed here.",
                     gumbo_normalized_tagname(error->input_tag));
       print_tag_stack(error, output);
       return;

suppressions/ruby.supp

@@ -98,7 +98,7 @@
   fun:evaluate
 }
 {
-  TODO
+  https://github.com/sparklemotion/nokogiri/actions/runs/5354163940/jobs/9710862134
   # 240 (120 direct, 120 indirect) bytes in 1 blocks are definitely lost in loss record 28,980 of 37,883
   # *xmlNewNodeEatName (tree.c:2299)
   # *xmlNewDocNodeEatName (tree.c:2374)
@@ -117,7 +117,7 @@
   fun:xmlNewNodeEatName
   fun:xmlNewDocNodeEatName
   fun:xmlSAX2StartElementNs
-  fun:xmlParseStartTag2
+  fun:xmlParseStartTag*
   fun:xmlParseElementStart
   fun:xmlParseContentInternal
   fun:xmlParseElement

test/helper.rb

@@ -128,7 +128,7 @@ class TestBenchmark < Minitest::BenchSpec
   end
 
   # rubocop:disable Style/ClassVars
-  class TestCase < MiniTest::Spec
+  class TestCase < Minitest::Spec
     include TestBase
 
     COMPACT_EVERY = 20
@@ -272,7 +272,7 @@ def assert_not_send(send_ary, m = nil)
     def pending(msg)
       begin
         yield
-      rescue MiniTest::Assertion
+      rescue Minitest::Assertion
         skip("pending #{msg} [#{caller(2..2).first}]")
       end
       flunk("pending test unexpectedly passed: #{msg} [#{caller(1..1).first}]")

test/test_memory_leak.rb

@@ -313,6 +313,21 @@ def test_object_space_memsize_of
     assert(bigger_name_size > base_size, "longer tags should increase memsize")
   end
 
+  def test_object_space_memsize_with_dtd
+    # https://github.com/sparklemotion/nokogiri/issues/2923
+    require "objspace"
+    skip("memsize_of not defined") unless ObjectSpace.respond_to?(:memsize_of)
+
+    doc = Nokogiri::XML(<<~XML)
+      <?xml version="1.0"?>
+      <!DOCTYPE staff PUBLIC "staff.dtd" [
+        <!ATTLIST payment type CDATA "check">
+      ]>
+      <staff></staff>
+    XML
+    ObjectSpace.memsize_of(doc) # assert_does_not_crash
+  end
+
   module MemInfo
     # from https://stackoverflow.com/questions/7220896/get-current-ruby-process-memory-usage
     # this is only going to work on linux