Fix being able to post URLs longer than 4096 characters (#17908)

app/lib/extractor.rb

@@ -1,6 +1,8 @@
 # frozen_string_literal: true
 
 module Extractor
+  MAX_DOMAIN_LENGTH = 253
+
   extend Twitter::TwitterText::Extractor
 
   module_function
@@ -30,6 +32,10 @@ module Extractor
       after      = $'
 
       unless Twitter::TwitterText::Regex[:end_mention_match].match?(after)
+        _, domain = screen_name.split('@')
+
+        next if domain.present? && domain.length > MAX_DOMAIN_LENGTH
+
         start_position = match_data.char_begin(1) - 1
         end_position   = match_data.char_end(1)
 

app/validators/status_length_validator.rb

@@ -3,35 +3,57 @@
 class StatusLengthValidator < ActiveModel::Validator
   MAX_CHARS = 500
   URL_PLACEHOLDER_CHARS = 23
-  URL_PLACEHOLDER = "\1#{'x' * URL_PLACEHOLDER_CHARS}"
+  URL_PLACEHOLDER = 'x' * 23
 
   def validate(status)
     return unless status.local? && !status.reblog?
 
-    @status = status
+    status.errors.add(:text, I18n.t('statuses.over_character_limit', max: MAX_CHARS)) if too_long?(status)
-    status.errors.add(:text, I18n.t('statuses.over_character_limit', max: MAX_CHARS)) if too_long?
   end
 
   private
 
-  def too_long?
+  def too_long?(status)
-    countable_length > MAX_CHARS
+    countable_length(combined_text(status)) > MAX_CHARS
   end
 
-  def countable_length
+  def countable_length(str)
-    total_text.mb_chars.grapheme_length
+    str.mb_chars.grapheme_length
   end
 
-  def total_text
+  def combined_text(status)
-    [@status.spoiler_text, countable_text].join
+    [status.spoiler_text, countable_text(status.text)].join
   end
 
-  def countable_text
+  def countable_text(str)
-    return '' if @status.text.nil?
+    return '' if str.blank?
 
-    @status.text.dup.tap do |new_text|
+    # To ensure that we only give length concessions to entities that
-      new_text.gsub!(FetchLinkCardService::URL_PATTERN, URL_PLACEHOLDER)
+    # will be correctly parsed during formatting, we go through full
-      new_text.gsub!(Account::MENTION_RE, '@\2')
+    # entity extraction
+
+    entities = Extractor.remove_overlapping_entities(Extractor.extract_urls_with_indices(str, extract_url_without_protocol: false) + Extractor.extract_mentions_or_lists_with_indices(str))
+
+    rewrite_entities(str, entities) do |entity|
+      if entity[:url]
+        URL_PLACEHOLDER
+      elsif entity[:screen_name]
+        "@#{entity[:screen_name].split('@').first}"
       end
     end
+  end
+
+  def rewrite_entities(str, entities)
+    entities.sort_by! { |entity| entity[:indices].first }
+    result = ''.dup
+
+    last_index = entities.reduce(0) do |index, entity|
+      result << str[index...entity[:indices].first]
+      result << yield(entity)
+      entity[:indices].last
+    end
+
+    result << str[last_index..-1]
+    result
+  end
 end

spec/validators/status_length_validator_spec.rb

@@ -50,6 +50,13 @@ describe StatusLengthValidator do
       expect(status.errors).to have_received(:add)
     end
 
+    it 'does not count overly long URLs as 23 characters flat' do
+      text = "http://example.com/valid?#{'#foo?' * 1000}"
+      status = double(spoiler_text: '', text: text, errors: double(add: nil), local?: true, reblog?: false)
+      subject.validate(status)
+      expect(status.errors).to have_received(:add)
+    end
+
     it 'counts only the front part of remote usernames' do
       text   = ('a' * 475) + " @alice@#{'b' * 30}.com"
       status = double(spoiler_text: '', text: text, errors: double(add: nil), local?: true, reblog?: false)
@@ -57,5 +64,13 @@ describe StatusLengthValidator do
       subject.validate(status)
       expect(status.errors).to_not have_received(:add)
     end
+
+    it 'does count both parts of remote usernames for overly long domains' do
+      text   = "@alice@#{'b' * 500}.com"
+      status = double(spoiler_text: '', text: text, errors: double(add: nil), local?: true, reblog?: false)
+
+      subject.validate(status)
+      expect(status.errors).to have_received(:add)
+    end
   end
 end