Merge commit '425d77f8124a50fc033e8fb3bdf7b89a6a25f4fa' into glitch-soc/merge-upstream

Conflicts:
- `.rubocop_todo.yml`:
  Upstream regenerated this file, glitch-soc had a specific ignore.
- `README.md`:
  Upstream updated its README, but glitch-soc has a completely different one.
  Kept glitch-soc's README

app/lib/importer/base_importer.rb

@@ -45,8 +45,11 @@ class Importer::BaseImporter
   # Remove documents from the index that no longer exist in the database
   def clean_up!
     index.scroll_batches do |documents|
+      primary_key = index.adapter.target.primary_key
+      raise ActiveRecord::UnknownPrimaryKey, index.adapter.target if primary_key.nil?
+
       ids           = documents.pluck('_id')
-      existence_map = index.adapter.target.where(id: ids).pluck(:id).each_with_object({}) { |id, map| map[id.to_s] = true }
+      existence_map = index.adapter.target.where(primary_key => ids).pluck(primary_key).each_with_object({}) { |id, map| map[id.to_s] = true }
       tmp           = ids.reject { |id| existence_map[id] }
 
       next if tmp.empty?

app/lib/request.rb

@@ -68,13 +68,26 @@ class Request
   # about 15s in total
   TIMEOUT = { connect_timeout: 5, read_timeout: 10, write_timeout: 10, read_deadline: 30 }.freeze
 
+  # Workaround for overly-eager decoding of percent-encoded characters in Addressable::URI#normalized_path
+  # https://github.com/sporkmonger/addressable/issues/366
+  URI_NORMALIZER = lambda do |uri|
+    uri = HTTP::URI.parse(uri)
+
+    HTTP::URI.new(
+      scheme: uri.normalized_scheme,
+      authority: uri.normalized_authority,
+      path: Addressable::URI.normalize_path(encode_non_ascii(uri.path)).presence || '/',
+      query: encode_non_ascii(uri.query)
+    )
+  end
+
   include RoutingHelper
 
   def initialize(verb, url, **options)
     raise ArgumentError if url.blank?
 
     @verb        = verb
-    @url         = Addressable::URI.parse(url).normalize
+    @url         = URI_NORMALIZER.call(url)
     @http_client = options.delete(:http_client)
     @allow_local = options.delete(:allow_local)
     @options     = options.merge(socket_class: use_proxy? || @allow_local ? ProxySocket : Socket)
@@ -138,8 +151,14 @@ class Request
       %w(http https).include?(parsed_url.scheme) && parsed_url.host.present?
     end
 
+    NON_ASCII_PATTERN = /[^\x00-\x7F]+/
+
+    def encode_non_ascii(str)
+      str&.gsub(NON_ASCII_PATTERN) { |substr| CGI.escape(substr.encode(Encoding::UTF_8)) }
+    end
+
     def http_client
-      HTTP.use(:auto_inflate).follow(max_hops: 3)
+      HTTP.use(:auto_inflate).use(normalize_uri: { normalizer: URI_NORMALIZER }).follow(max_hops: 3)
     end
   end
 

app/lib/rss/builder.rb

@@ -14,13 +14,14 @@ class RSS::Builder
   end
 
   def to_xml
-    ('<?xml version="1.0" encoding="UTF-8"?>'.dup << Ox.dump(wrap_in_document, effort: :tolerant)).force_encoding('UTF-8')
+    Ox.dump(wrap_in_document, effort: :tolerant).force_encoding('UTF-8')
   end
 
   private
 
   def wrap_in_document
     Ox::Document.new(version: '1.0').tap do |document|
+      document << xml_instruct
       document << Ox::Element.new('rss').tap do |rss|
         rss['version']        = '2.0'
         rss['xmlns:webfeeds'] = 'http://webfeeds.org/rss/1.0'
@@ -30,4 +31,11 @@ class RSS::Builder
       end
     end
   end
+
+  def xml_instruct
+    Ox::Instruct.new(:xml).tap do |instruct|
+      instruct[:version] = '1.0'
+      instruct[:encoding] = 'UTF-8'
+    end
+  end
 end

app/lib/text_formatter.rb

@@ -75,7 +75,7 @@ class TextFormatter
       entity[:indices].first
     end
 
-    result = ''.dup
+    result = +''
 
     last_index = entities.reduce(0) do |index, entity|
       indices = entity[:indices]