require "helper" class TestMemoryLeak < Nokogiri::TestCase def setup super @str = <
EOF end # # this suite is turned off unless the env var NOKOGIRI_GC is non-nil # # to run any of these tests, do something like this on the commandline: # # $ NOKOGIRI_GC=t ruby -Ilib:test \ # test/test_memory_leak.rb \ # -n /test_leaking_namespace_node_strings/ # # also see: # # https://github.com/sparklemotion/nokogiri/issues/1603 # # which is an open issue to resurrect these tests and run them as # part of the CI pipeline. # if ENV['NOKOGIRI_GC'] # turning these off by default for now def test_dont_hurt_em_why content = File.open("#{File.dirname(__FILE__)}/files/dont_hurt_em_why.xml").read ndoc = Nokogiri::XML(content) 2.times do ndoc.search('status text').first.inner_text ndoc.search('user name').first.inner_text GC.start end end class BadIO def read(*args) raise 'hell' end def write(*args) raise 'chickens' end end def test_for_mem_leak_on_io_callbacks io = File.open SNUGGLES_FILE Nokogiri::XML.parse(io) loop do Nokogiri::XML.parse(BadIO.new) rescue nil doc.write BadIO.new rescue nil end end def test_for_memory_leak begin # we don't use Dike in any tests, but requiring it has side effects # that can create memory leaks, and that's what we're testing for. require 'rubygems' require 'dike' # do not remove! count_start = count_object_space_documents xml_data = <<-EOS abc 1234 Zzz EOS 20.times do doc = Nokogiri::XML(xml_data) doc.xpath("//item") end 2.times { GC.start } count_end = count_object_space_documents assert((count_end - count_start) <= 2, "memory leak detected") rescue LoadError puts "\ndike is not installed, skipping memory leak test" end end def test_node_set_namespace_mem_leak xml = Nokogiri::XML "" ctx = Nokogiri::XML::XPathContext.new(xml) loop do ctx.evaluate("//namespace::*") end end def test_leak_on_node_replace loop do doc = Nokogiri.XML("") n = Nokogiri::XML::CDATA.new(doc, "bar") pivot = doc.root.children[0] pivot.replace(n) end end def test_sax_parser_context io = StringIO.new(@str) loop do Nokogiri::XML::SAX::ParserContext.new(@str) Nokogiri::XML::SAX::ParserContext.new(io) io.rewind Nokogiri::HTML::SAX::ParserContext.new(@str) Nokogiri::HTML::SAX::ParserContext.new(io) io.rewind end end class JumpingSaxHandler < Nokogiri::XML::SAX::Document def initialize(jumptag) @jumptag = jumptag super() end def start_element(name, attrs = []) throw @jumptag end end def test_jumping_sax_handler doc = JumpingSaxHandler.new(:foo) loop do catch(:foo) do Nokogiri::HTML::SAX::Parser.new(doc).parse(@str) end end end def test_in_context_parser_leak loop do doc = Nokogiri::XML::Document.new fragment1 = Nokogiri::XML::DocumentFragment.new(doc, '') node = fragment1.children[0] node.parse('') end end def test_in_context_parser_leak_ii loop { Nokogiri::XML('').root.parse('') } end def test_leak_on_xpath_string_function doc = Nokogiri::XML(@str) loop do doc.xpath('name(//node())') end end def test_leaking_namespace_node_strings # see https://github.com/sparklemotion/nokogiri/issues/1810 for memory leak report ns = {'xmlns' => 'http://schemas.xmlsoap.org/soap/envelope/'} 20.times do 10_000.times do Nokogiri::XML::Builder.new do |xml| xml.send 'Envelope', ns do xml.send 'Foobar', ns end end end puts MemInfo.rss end end def test_leaking_namespace_node_strings_with_prefix # see https://github.com/sparklemotion/nokogiri/issues/1810 for memory leak report ns = {'xmlns:foo' => 'http://schemas.xmlsoap.org/soap/envelope/'} 20.times do 10_000.times do Nokogiri::XML::Builder.new do |xml| xml.send 'Envelope', ns do xml.send 'Foobar', ns end end end puts MemInfo.rss end end def test_leaking_dtd_nodes_after_internal_subset_removal # see https://github.com/sparklemotion/nokogiri/issues/1784 100_000.times do |i| doc = Nokogiri::HTML::Document.new doc.internal_subset.remove puts MemInfo.rss if (i % 1000 == 0) end end describe "#2114 RelaxNG schema parsing has a small memory leak" do it "no longer leaks" do prev_rss = MemInfo.rss 100_001.times do |j| Nokogiri::XML::RelaxNG.from_document(Nokogiri::XML::Document.parse(File.read(ADDRESS_SCHEMA_FILE))) if (j % 10_000 == 0) curr_rss = MemInfo.rss diff_rss = curr_rss - prev_rss printf("\n(iter %d) %d", j, curr_rss) printf(" (%s%d)", diff_rss >= 0 ? "+" : "-", diff_rss) if j > 0 prev_rss = curr_rss end end puts end end end # if NOKOGIRI_GC module MemInfo # from https://stackoverflow.com/questions/7220896/get-current-ruby-process-memory-usage # this is only going to work on linux PAGE_SIZE = `getconf PAGESIZE`.chomp.to_i rescue 4096 STATM_PATH = "/proc/#{Process.pid}/statm" STATM_FOUND = File.exist?(STATM_PATH) def self.rss if STATM_FOUND return (File.read(STATM_PATH).split(' ')[1].to_i * PAGE_SIZE) / 1024 end return 0 end end private def count_object_space_documents count = 0 ObjectSpace.each_object {|j| count += 1 if j.is_a?(Nokogiri::XML::Document) } count end end