test/ruby/test_io.rb (test_recycled_fd_close): Linux workaround

Arch Linux CI still seems to timeout on this test...

Note, I can't reproduce the failures in these tests on a
FreeBSD 11.1 VM while infinite-looping, even without the
"th.join(0.001)".  It doesn't seem related to the use of
rb_wait_for_single_fd (r64529).

cf. https://rubyci.org/logs/rubyci.s3.amazonaws.com/arch/ruby-trunk/log/20180826T090003Z.fail.html.gz

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64543 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

1 files changed, 12 insertions, 0 deletions

diff --git a/test/ruby/test_io.rb b/test/ruby/test_io.rb
index f2e9128e8a..7843fa0877 100644
--- a/test/ruby/test_io.rb
+++ b/test/ruby/test_io.rb
@@ -3781,6 +3781,15 @@ __END__
           IO.pipe do |r, w|
             th = Thread.new { r.read(1) }
             w.write(dot)
+
+            # XXX not sure why this is needed on Linux, otherwise
+            # the "good" reader thread doesn't always join properly
+            # because the reader never sees the first write
+            if RUBY_PLATFORM =~ /linux/
+              # assert_equal can fail if this is another char...
+              w.write(dot)
+            end
+
             assert_same th, th.join(15), '"good" reader timeout'
             assert_equal(dot, th.value)
           end
@@ -3798,7 +3807,10 @@ __END__
             end
           end
           Thread.pass until th.stop?
+
+          # XXX not sure why, this reduces Linux CI failures
           assert_nil th.join(0.001)
+
           r.close
           assert_same th, th.join(30), '"bad" reader timeout'
           assert_match(/stream closed/, th.value.message)