RE: [PATCH] Fix sporadic failure in g++.dg/tsan/aligned_vs_unaligned_race.C

[Bernd Edlinger <bernd.edlinger@hotmail.de>]

[-- Attachment #1: Type: text/plain, Size: 1038 bytes --]

Aehm, sorry,

that's the sporadic failure, I mentioned:

https://gcc.gnu.org/ml/gcc-regression/2015-01/msg00041.html
New failures:
FAIL: g++.dg/tsan/aligned_vs_unaligned_race.C   -O2  execution test

New passes:

https://gcc.gnu.org/ml/gcc-regression/2015-01/msg00044.html
New failures:

New passes:
FAIL: g++.dg/tsan/aligned_vs_unaligned_race.C   -O2  execution test

Really I did run this test often, before I checked it in, but...

It is due to a race condition in tsan itself, it cannot decide which access was the
previous one and which was the second one, but our tsan tests are not meant as
a functional test of the tsan runtime library, they are only meant to test the GCC
instrumentation.

For the purpose of finding race conditions in an application a detection likelihood of 98%
is absolutely perfect, just for our test suite that causes unnecessary failures.


So, I still think I should fix that test case, maybe with a comment, why I have to
sleep(1), what do you think?


Bernd.
 		 	   		  

[-- Attachment #2: patch-tsan-race.diff --]
[-- Type: application/octet-stream, Size: 718 bytes --]

2015-01-03  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * g++.dg/tsan/aligned_vs_unaligned_race.C: Fixed sporadic failure.


Index: gcc/testsuite/g++.dg/tsan/aligned_vs_unaligned_race.C
===================================================================
--- gcc/testsuite/g++.dg/tsan/aligned_vs_unaligned_race.C	(revision 219160)
+++ gcc/testsuite/g++.dg/tsan/aligned_vs_unaligned_race.C	(working copy)
@@ -2,10 +2,14 @@
 #include <pthread.h>
 #include <stdio.h>
 #include <stdint.h>
+#include <unistd.h>
 
 uint64_t Global[2];
 
 void *Thread1(void *x) {
+  /* We have to sleep here, to make it somewhat easier for tsan to
+     detect the race condition.  */
+  sleep(1);
   Global[1]++;
   return NULL;
 }