Alex
Alex
There are a lot of true positives (especially, in `tbbmalloc`). We are fixing them from time to time. Some basics tests, (e.g. for `parallel_for`, `parallel_reduce` and `task_group`) are not expected...
> I would have thought threading correctness issues in a threading library would be high priority. :) In my experience, TSan makes these otherwise hard to debug and hard to...
It seems like a new issue ; however, this feature is not released yet. If possible, can you please try the previous revision: https://github.com/oneapi-src/oneTBB/tree/b15aabb39a12a4dcddb6ee886265f2bf2f972650? notify @pavelkumbrasev
It turned out that Thread Sanitizer does not recognize synchronization via semaphores. The following code reports a race: ``` #include #include #include #include #include #include class binary_semaphore { public: binary_semaphore()...
I am also not an expert in mach semaphores but at first glance the example is correct (at least TBB is based on this implementation for many years). So, it...
> it would be nice if TBB had an "official" suppression list of known issues. Yes, it makes sense.
Good catch! Thank you for the investigation. Can you try the workaround? Try to add backoff strategy to the `pop` method ([task_stream.h:196-199](https://github.com/oneapi-src/oneTBB/blob/master/src/tbb/task_stream.h#L196-L199)): ```diff - do { - lane = next_lane(...
> The behavior doesn't change. Yes, it seems another issue. The observed assert is about relation between `market` and `arena` while #739 fixes an issue in `private_server` (thread pool).
It seems the testing approach is broken on ARM, any test using `utils::SpinBarrier` might hang without real issue. We are thinking about better approach for testing.