Long story short. Tsan has a logical data race the core of data race detection algorithm. The race is not a bug, but a deliberate design decision that makes tsan considerably faster. So ironically, if the race memory accesses happen almost simultaneously, tsan can miss the race. Thus we have sleeps. Sleeps vs barrier is being discussed in the "Fix parameters of __tsan_vptr_update" thread. I would really like to keep llvm and gcc tests in sync as much as possible.
On Fri, Jan 9, 2015 at 6:36 PM, Jakub Jelinek <ja...@redhat.com> wrote: > On Fri, Jan 09, 2015 at 04:32:47PM +0100, Bernd Edlinger wrote: >> Hi, >> >> On Thu, 8 Jan 2015 22:27:26, Jakub Jelinek wrote: >> >> Any objections to approving it now? >> > >> > LGTM. >> > >> > Jakub >> >> would it be OK to apply this patch also to the 4.9 testsuite, >> except for c-c++-common/tsan/bitfield_race.c and >> g++.dg/tsan/aligned_vs_unaligned_race.C of course? > > Yes, but please give Dmitry some time to respond. > > Jakub