(v1.11.x) update tests to work with system libxml 2.9.12, and work around windows dll unloading issue

.github/workflows/windows.yml

@@ -0,0 +1,50 @@
+# this is a work in progress!
+name: windows
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    types: [opened, synchronize]
+    branches:
+      - '*'
+
+jobs:
+  windows:
+    name: "windows, sys: ${{ matrix.sys }}, ${{ matrix.ruby }}"
+
+    env:
+      MAKEFLAGS: -j2
+
+    runs-on: windows-latest
+
+    strategy:
+      fail-fast: false
+      matrix:
+        sys: [ enable, disable ]
+        ruby: [ "2.5", "2.6", "2.7", "3.0", "mingw" ]
+
+    steps:
+      - name: configure git crlf on windows
+        run: |
+          git config --system core.autocrlf false
+          git config --system core.eol lf
+      - name: checkout
+        uses: actions/checkout@v2
+      - name: load Ruby and bundle install
+        uses: MSP-Greg/setup-ruby-pkgs@v1
+        with:
+          ruby-version: ${{ matrix.ruby }}
+          mingw: libxml2 libxslt
+          bundler-cache: true
+      - uses: actions/cache@v2
+        if: matrix.sys == 'disable'
+        with:
+          path: ports/archives
+          key: ${{ matrix.os }}-${{ matrix.ruby }}-tarballs-${{ hashFiles('**/dependencies.yml') }}
+          restore-keys: ${{ matrix.os }}-${{ matrix.ruby }}-tarballs-
+      - name: bundle exec rake compile
+        run: |
+          bundle exec rake compile -- --${{ matrix.sys }}-system-libraries
+      - name: bundle exec rake test
+        run:  bundle exec rake test

CHANGELOG.md

@@ -4,6 +4,22 @@ Nokogiri follows [Semantic Versioning](https://semver.org/), please see the [REA
 
 ---
 
+## 1.11.5 / 2021-05-19
+
+### Fixed
+
+[Windows CRuby] Work around segfault at process exit on Windows when using libxml2 system DLLs.
+
+libxml 2.9.12 introduced new behavior to avoid memory leaks when unloading libxml2 shared libraries (see [libxml/!66](https://gitlab.gnome.org/GNOME/libxml2/-/merge_requests/66)). Early testing caught this segfault on non-Windows platforms (see [#2059](https://github.com/sparklemotion/nokogiri/issues/2059) and [libxml@956534e](https://gitlab.gnome.org/GNOME/libxml2/-/commit/956534e02ef280795a187c16f6ac04e107f23c5d)) but it was incompletely fixed and is still an issue on Windows platforms that are using system DLLs.
+
+We work around this by configuring libxml2 in this situation to use its default memory management functions. Note that if Nokogiri is not on Windows, or is not using shared system libraries, it will will continue to configure libxml2 to use Ruby's memory management functions. `Nokogiri::VERSION_INFO["libxml"]["memory_management"]` will allow you to verify when the default memory management functions are being used. [[#2241](https://github.com/sparklemotion/nokogiri/issues/2241)]
+
+
+### Changed
+
+`Nokogiri::VERSION_INFO["libxml"]` now contains the key `"memory_management"` to declare whether libxml2 is using its `default` memory management functions, or whether it uses the memory management functions from `ruby`. See above for more details.
+
+
 ## 1.11.4 / 2021-05-14
 
 ### Security

ext/nokogiri/nokogiri.c

@@ -191,7 +191,26 @@ Init_nokogiri()
   rb_const_set(mNokogiri, rb_intern("OTHER_LIBRARY_VERSIONS"), NOKOGIRI_STR_NEW2(NOKOGIRI_OTHER_LIBRARY_VERSIONS));
 #endif
 
+#if defined(_WIN32) && !defined(NOKOGIRI_PACKAGED_LIBRARIES)
+  /*
+   *  We choose *not* to do use Ruby's memory management functions with windows DLLs because of this
+   *  issue in libxml 2.9.12:
+   *
+   *    https://github.com/sparklemotion/nokogiri/issues/2241
+   *
+   *  If the atexit() issue gets fixed in a future version of libxml2, then we may be able to skip
+   *  this config only for the specific libxml2 versions 2.9.12.
+   *
+   *  Alternatively, now that Ruby has a generational GC, it might be OK to let libxml2 use its
+   *  default memory management functions (recall that this config was introduced to reduce memory
+   *  bloat and allow Ruby to GC more often); but we should *really* test with production workloads
+   *  before making that kind of a potentially-invasive change.
+   */
+  rb_const_set(mNokogiri, rb_intern("LIBXML_MEMORY_MANAGEMENT"), NOKOGIRI_STR_NEW2("default"));
+#else
+  rb_const_set(mNokogiri, rb_intern("LIBXML_MEMORY_MANAGEMENT"), NOKOGIRI_STR_NEW2("ruby"));
   xmlMemSetup((xmlFreeFunc)ruby_xfree, (xmlMallocFunc)ruby_xmalloc, (xmlReallocFunc)ruby_xrealloc, ruby_strdup);
+#endif
 
   xmlInitParser();
 

lib/nokogiri/version/info.rb

@@ -137,6 +137,7 @@ def to_hash
             else
               libxml["source"] = "system"
             end
+            libxml["memory_management"] = Nokogiri::LIBXML_MEMORY_MANAGEMENT
             libxml["iconv_enabled"] = libxml2_has_iconv?
             libxml["compiled"] = compiled_libxml_version.to_s
             libxml["loaded"] = loaded_libxml_version.to_s

test/html/test_comments.rb

@@ -113,8 +113,7 @@ class TestComment < Nokogiri::TestCase
         let(:subject) { doc.at_css("div#under-test") }
         let(:inner_div) { doc.at_css("div#do-i-exist") }
 
-        if Nokogiri.uses_libxml? && Nokogiri::VersionInfo.instance.libxml2_using_packaged?
-          # see patches/libxml2/0006-htmlParseComment-treat-as-if-it-closed-the-comment.patch
+        if Nokogiri::VersionInfo.instance.libxml2_using_packaged? || (Nokogiri::VersionInfo.instance.libxml2_using_system? && Nokogiri.uses_libxml?(">=2.9.11"))
           it "behaves as if the comment is normally closed" do # COMPLIANT
             assert_equal 3, subject.children.length
             assert subject.children[0].comment?
@@ -128,9 +127,7 @@ class TestComment < Nokogiri::TestCase
           end
         end
 
-        if Nokogiri.jruby? || Nokogiri::VersionInfo.instance.libxml2_using_system?
-          # this behavior may change to the above in libxml v2.9.11 depending on whether
-          # https://gitlab.gnome.org/GNOME/libxml2/-/merge_requests/82 is merged
+        if Nokogiri.jruby? || (Nokogiri::VersionInfo.instance.libxml2_using_system? && Nokogiri.uses_libxml?("<2.9.11"))
           it "behaves as if the comment encompasses the inner div" do # NON-COMPLIANT
             assert_equal 1, subject.children.length
             assert subject.children.first.comment?