Remove unavailable characters when encoding page title to ISO-8859-1tc-page-title-encoding-fix

The Page-Title header has to be ISO-8859-1, so when converting the
title to that encoding, remove characters that are not available.

Closes gitlab-org/gitlab-ce#39179.

3 files changed, 25 insertions, 1 deletions

diff --git a/app/controllers/application_controller.rb b/app/controllers/application_controller.rb
index 967fe39256a..505f8c2f5f6 100644
--- a/app/controllers/application_controller.rb
+++ b/app/controllers/application_controller.rb
@@ -349,6 +349,6 @@ class ApplicationController < ActionController::Base
 
   def set_page_title_header
     # Per https://tools.ietf.org/html/rfc5987, headers need to be ISO-8859-1, not UTF-8
-    response.headers['Page-Title'] = page_title('GitLab').encode('ISO-8859-1')
+    response.headers['Page-Title'] = Gitlab::EncodingHelper.lossy_encode(page_title('GitLab'), 'ISO-8859-1')
   end
 end
diff --git a/lib/gitlab/encoding_helper.rb b/lib/gitlab/encoding_helper.rb
index 7b3483a7f96..0efeecd7d4a 100644
--- a/lib/gitlab/encoding_helper.rb
+++ b/lib/gitlab/encoding_helper.rb
@@ -64,6 +64,11 @@ module Gitlab
       end
     end
 
+    # Convert all characters to the desired encoding, and drop the characters not available in that encoding
+    def lossy_encode(message, encoding)
+      message.encode(encoding, invalid: :replace, undef: :replace, replace: '')
+    end
+
     private
 
     def clean(message)
diff --git a/spec/lib/gitlab/encoding_helper_spec.rb b/spec/lib/gitlab/encoding_helper_spec.rb
index 8b14b227e65..1838b21c5a2 100644
--- a/spec/lib/gitlab/encoding_helper_spec.rb
+++ b/spec/lib/gitlab/encoding_helper_spec.rb
@@ -1,6 +1,8 @@
 require "spec_helper"
 
 describe Gitlab::EncodingHelper do
+  using RSpec::Parameterized::TableSyntax
+
   let(:ext_class) { Class.new { extend Gitlab::EncodingHelper } }
   let(:binary_string) { File.read(Rails.root + "spec/fixtures/dk.png") }
 
@@ -109,6 +111,23 @@ describe Gitlab::EncodingHelper do
     end
   end
 
+  describe '#lossy_encode' do
+    context 'to ISO-8859-1' do
+      where(:language, :input, :output) do
+        'ASCII'     | 'plain ASCII'          | 'plain ASCII'
+        'Portugese' | 'Última Atualização'   | 'Última Atualização'
+        'Russian'   | 'Последнее обновление' | ' '
+        'Esperanto' | 'Lasta ĝisdatigo'      | 'Lasta isdatigo'
+      end
+
+      with_them do
+        it 'properly encodes' do
+          expect(ext_class.lossy_encode(input, 'ISO-8859-1').encode('UTF-8')).to eq(output)
+        end
+      end
+    end
+  end
+
   describe '#clean' do
     [
       [