Unit test lf-t unstable

[Laurynas Biveinis]

11/14 Test #11: lf ...............................***Failed 7.12 sec
# N CPUs: 4, atomic ops: gcc-builtins-smp
1..7
ok 1 - my_atomic_initialize() returned 0
# Testing lf_pinbox (with my_thread_init) with 30 threads, 3000 iterations...
ok 2 - tested lf_pinbox (with my_thread_init) in 0.0131241 secs (0)
# Testing lf_alloc (with my_thread_init) with 30 threads, 3000 iterations...
# 0 mallocs, 4 pins in stack
ok 3 - tested lf_alloc (with my_thread_init) in 0.0163577 secs (0)
# Testing lf_hash (with my_thread_init) with 30 threads, 300 iterations...
# 0 mallocs, 2 pins in stack, 1 hash size, 0 inserts
ok 4 - tested lf_hash (with my_thread_init) in 0.0053816 secs (0)
# Testing lf_pinbox (without my_thread_init) with 30 threads, 3000 iterations...
ok 5 - tested lf_pinbox (without my_thread_init) in 0.0096111 secs (0)
# Testing lf_alloc (without my_thread_init) with 30 threads, 3000 iterations...
# 0 mallocs, 5 pins in stack
ok 6 - tested lf_alloc (without my_thread_init) in 0.0579291 secs (0)
# Testing lf_hash (without my_thread_init) with 30 threads, 300 iterations...
# 0 mallocs, 2 pins in stack, 1 hash size, 0 inserts
ok 7 - tested lf_hash (without my_thread_init) in 0.0042839 secs (0)
Error in my_thread_global_end(): 1 threads didn't exit

=================================================================
==1810==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 280 byte(s) in 1 object(s) allocated from:
    #0 0x7fb21177579a in __interceptor_calloc (/usr/lib/x86_64-linux-gnu/libasan.so.2+0x9879a)
    #1 0x4089cc in my_thread_init /mnt/workspace/percona-server-5.5-asan-param/BUILD_TYPE/debug-asan/Host/ubuntu-xenial-64bit/mysys/my_thr_init.c:344
    #2 0x40321e in test_lf_hash /mnt/workspace/percona-server-5.5-asan-param/BUILD_TYPE/debug-asan/Host/ubuntu-xenial-64bit/unittest/mysys/lf-t.c:124
    #3 0x7fb2114c76f9 in start_thread (/lib/x86_64-linux-gnu/libpthread.so.0+0x76f9)

SUMMARY: AddressSanitizer: 280 byte(s) leaked in 1 allocation(s).

This is caused by a race: the last test thread with my_thread_init=1 reads with_my_thread_init 2nd time to determine whether it should call my_thread_end, after signaling the main thread that the test is complete. And the main thread sets my_thread_init from 1 to 0 after the first three tests, resulting in my_thread_init/my_thread_end mismatch.

[Laurynas Biveinis (laurynas-biveinis)]

     https://github.com/percona/percona-server/pull/776
     https://github.com/percona/percona-server/pull/777
     https://github.com/percona/percona-server/pull/778

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-2165