AO3-5826 Update nokogiri to 1.10.5 (#3697)

* [Security] Bump nokogiri from 1.10.4 to 1.10.5

Bumps [nokogiri](https://github.com/sparklemotion/nokogiri) from 1.10.4 to 1.10.5. **This update includes a security fix.**
- [Release notes](https://github.com/sparklemotion/nokogiri/releases)
- [Changelog](https://github.com/sparklemotion/nokogiri/blob/master/CHANGELOG.md)
- [Commits](sparklemotion/nokogiri@v1.10.4...v1.10.5)

Signed-off-by: dependabot-preview[bot] <[email protected]>

* Remove code for closing uncode tags that relied on an error message that no longer exists in libxml2

* Test that shows contents of third paragraph are not as expected

* Remove code for closing uncode tags that relied on an error message that no longer exists in libxml2

* ACTUALLY the test that shows the contents of the third paragraph

* Third time is the charm showing the useful test failure

* Write a passing test and add a comment explaining it does not really resemble reality

Gemfile.lock

@@ -261,7 +261,7 @@ GEM
     netrc (0.11.0)
     newrelic_rpm (6.4.0.356)
     nio4r (2.3.1)
-    nokogiri (1.10.4)
+    nokogiri (1.10.5)
       mini_portile2 (~> 2.4.0)
     nokogumbo (1.4.9)
       nokogiri

lib/html_cleaner.rb

@@ -354,15 +354,8 @@ def close_unclosed_tag(text, tag, line_number)
   end
 
   def add_paragraphs_to_text(text)
-    # By default, Nokogiri closes unclosed tags very late, often at
-    # the end of the document. We want runaway tags closed at the end
-    # of the line
     doc = Nokogiri::XML.parse("<myroot>#{text}</myroot>")
     doc.errors.each do |error|
-      match = error.message.match(/Premature end of data in tag (\w+) line (\d+)/)
-
-      text = close_unclosed_tag(text, match[1], match[2]) if match
-
       match = error.message.match(/Opening and ending tag mismatch: (\w+) line (\d+) and myroot/)
       text = close_unclosed_tag(text, match[1], match[2]) if match
     end

spec/miscellaneous/lib/html_cleaner_spec.rb

@@ -845,7 +845,15 @@
       expect(doc.xpath("./p[contains(@class, 'bar')]").children.to_s.strip).to eq("foobar")
     end
 
-    it "should close unclosed inline tags before double linebreak" do
+    # When we call add_paragraphs_to_text, everything gets wrapped inside myroot
+    # tags, and the closing myroot tag is treated as a mismatch for strong, ergo
+    # strong is closed on the second paragraph while the em tag remains open.
+    # In real world use, however, this content would most likely be run through
+    # Sanitize.clean first, which would close both the em and strong tags at the
+    # very end, so we wouldn't have a mismatch and the strong tag would be
+    # reopened in every paragraph, just like the em tag is. More info at:
+    # https://github.com/otwcode/otwarchive/pull/3692#issuecomment-558740913
+    it "should close mismatched tags" do
       html = """Here is an unclosed <em>em tag.
 
       Here is an unclosed <strong>strong tag.
@@ -854,11 +862,11 @@
 
       doc = Nokogiri::HTML.fragment(add_paragraphs_to_text(html))
       expect(doc.xpath("./p[1]/em").children.to_s.strip).to eq("em tag.")
-      expect(doc.xpath("./p[2]/strong").children.to_s.strip).to eq("strong tag.")
-      expect(doc.xpath("./p[3]").children.to_s.strip).to eq("Stuff.")
+      expect(doc.xpath("./p[2]/em/strong").children.to_s.strip).to eq("strong tag.")
+      expect(doc.xpath("./p[3]/em").children.to_s.strip).to eq("Stuff.")
     end
 
-    it "should close unclosed tag withing other tag" do
+    it "should close unclosed tag within other tag" do
       pending "Opened bug report with Nokogiri"
       html = "<strong><em>unclosed</strong>"
       doc = Nokogiri::HTML.fragment(add_paragraphs_to_text(html))