Add tsan suppression for gamepad unittest code

tools/valgrind/tsan/suppressions.txt

 # There are a few kinds of suppressions in this file.
 # 1. third party stuff we have no control over
 #
 # 2. intentional unit test errors, or stuff that is somehow a false positive
 # in our own code, or stuff that is so trivial it's not worth fixing
 #
 # 3. Suppressions for real chromium bugs that are not yet fixed.
 # These should all be in chromium's bug tracking system (but a few aren't yet).
 # Periodically we should sweep this file and the bug tracker clean by
 # running overnight and removing outdated bugs/suppressions.
@@ skipping 374 matching lines @@
   fun:base::Histogram::*
   fun:HistogramSynchronizer::DeserializeHistogramList*
 }
 
 {
    bug_104776 Benign race to initialize pointer with the same value again.
    ThreadSanitizer:Race
    fun:base::StatisticsRecorder::FindHistogram
    fun:base::*Histogram::FactoryGet
 }
+{
+  bug_79050 Test to detect read contention & benign race
+  ThreadSanitizer:Race

[dhollowa, 2011/11/22 21:35:09]

Why did you choose to elide the lines?:

   fun:Replace_memcpy
   fun:memcpy

[Timur Iskhodzhanov, 2011/11/25 10:18:57]

Actually, the eliding would help:
http://build.chromium.org/p/chromium.memory.fyi/builders/Linux%20Tests%20%28t...
WARNING: Possible data race during read of size 8 at 0x45D8008: {{{
   T83 (Gamepad polling thread) (L{}):
    #0  ::MockDataFetcher::GetGamepadData
content/browser/gamepad/gamepad_provider_unittest.cc:22
    #1  gamepad::Provider::DoPoll
content/browser/gamepad/gamepad_provider.cc:123
    #2  base::internal::RunnableAdapter::Run base/bind_internal.h:140
    #3  base::internal::InvokeHelper::MakeItSo base/bind_internal.h:771
    #4  base::internal::Invoker::Run base/bind_internal.h:1010
    #5  base::Callback::Run base/callback.h:268
    #6  MessageLoop::RunTask base/message_loop.cc:500
  Concurrent write(s) happened at (OR AFTER) these points:
   T0 (L{}):
    #0  ::MockDataFetcher::SetData
content/browser/gamepad/gamepad_provider_unittest.cc:26
    #1  ::GamepadProviderTest_FLAKY_PollingAccess_Test::TestBody
content/browser/gamepad/gamepad_provider_unittest.cc:82
    #2  testing::internal::HandleSehExceptionsInMethodIfSupported
testing/gtest/src/gtest.cc:2090
  Location 0x45D8008 is 8 bytes inside a block starting at 0x45D8000 of size
1656 allocated by T0 from heap:
    #0  operator new /tmp/valgrind-src/tsan/tsan/ts_valgrind_intercepts.c:426
    #1  ::GamepadProviderTest::CreateProvider
content/browser/gamepad/gamepad_provider_unittest.cc:44
    #2  ::GamepadProviderTest_FLAKY_PollingAccess_Test::TestBody
content/browser/gamepad/gamepad_provider_unittest.cc:70
    #3  testing::internal::HandleSehExceptionsInMethodIfSupported
testing/gtest/src/gtest.cc:2090
   Race verifier data: 0x453486,0x4530BA
}}}

Rationale for dropping lines like memcpy for TSan:
a) the code lacks synchronization.
hence
b) writes below GetGamepadData may race with something else.
memcpy is a form of write, right?

On 2011/11/22 21:35:09, dhollowa wrote:

+  ...
+  fun:::MockDataFetcher::GetGamepadData
+  fun:gamepad::Provider::DoPoll
+}
 
 # 3. Suppressions for real chromium bugs that are not yet fixed.
 ############################
 # Real races in Chromium
 {
   bug_23433
   ThreadSanitizer:Race
   fun:*logging*SetMinLogLevel*
 }
 
@@ skipping 268 matching lines @@
 {
    bug_104769
    ThreadSanitizer:Race
    fun:timeout_correct
    fun:event_base_loop
    fun:base::MessagePumpLibevent::Run
    fun:MessageLoop::RunInternal
    fun:MessageLoop::RunHandler
    fun:MessageLoop::Run
 }