bzr push into mysql-5.6 branch (tor.didriksen:4497 to 4498) Bug#14689561

[Tor Didriksen]

 4498 Tor Didriksen	2012-10-17
      Bug#14689561 - SERVER CRASH ON CONNECTING CLIENT 
      
      Post-push fix: We must call hash_filo::clear() to initialize
      the mutex, before calling resize(0).

    modified:
      unittest/gunit/segfault-t.cc
 4497 Tor Didriksen	2012-10-17
      Bug#14771006 SHOW PROCESSLIST SORT ORDER CHANGED IN 5.6
      
      'SHOW PROCESSLIST' has traditionally returned the result ordered
      by thread_id, so do the same now, even if the THD-list is a std::set.
      
      Solution:
      Convert thread_infos from intrusive list, to array of pointers, and std::sort it.

    modified:
      sql/global_threads.h
      sql/sql_show.cc
=== modified file 'unittest/gunit/segfault-t.cc'
--- a/unittest/gunit/segfault-t.cc	2012-10-16 08:39:32 +0000
+++ b/unittest/gunit/segfault-t.cc	2012-10-17 15:11:15 +0000
@@ -180,16 +180,19 @@ TEST(PrintUtilities, Printf)
   EXPECT_STREQ(sprintfbuff, buff);
 }
 
-TEST_F(FatalSignalDeathTest, TestHashFiloZeroSize)
+
+// After the fix for Bug#14689561, this is no longer a death test.
+TEST(HashFiloTest, TestHashFiloZeroSize)
 {
   hash_filo *t_cache;
   t_cache= new hash_filo(5, 0, 0,
                          (my_hash_get_key) NULL,
                          (my_hash_free_key) NULL,
                          NULL);
+  t_cache->clear();
   t_cache->resize(0);
   hash_filo_element entry;
-  //after resize (to zero) it tries to dereference last_link which is NULL
+  // After resize (to zero) it tries to dereference last_link which is NULL.
   t_cache->add(&entry);
   delete t_cache;
 }

No bundle (reason: useless for push emails).