False positive from clang-analyzer-deadcode.DeadStores with reference and thread

[Quuxplusone]

Bugzilla Link	PR51674
Status	NEW
Importance	P enhancement
Reported by	Sam Kearney (samuelmkearney@gmail.com)
Reported on	2021-08-30 07:50:34 -0700
Last modified on	2021-09-03 09:27:46 -0700
Version	unspecified
Hardware	PC Windows NT
CC	dcoughlin@apple.com, llvm-bugs@lists.llvm.org, nok.raven@gmail.com
Fixed by commit(s)
Attachments
Blocks
Blocked by
See also

To reproduce: Install LLVM 14.0, add this minimal example as main.cpp:

#include <chrono>
#include <thread>
#include <iostream>

using namespace std::chrono_literals;

int main()
{
  bool keep_running_thread = true;

  std::thread thrd(
      [](const bool& keep_running) {
        while (keep_running)
        {
          std::cout << "Still running!\n";
          std::this_thread::sleep_for(200ms);
        }
      },
      std::ref(keep_running_thread));

  std::this_thread::sleep_for(1s);
  keep_running_thread = false;
  thrd.join();
  return 0;
}

Run clang-tidy --checks=clang-analyzer-* main.cpp. It reports:

main.cpp:22:3: warning: Value stored to 'keep_running_thread' is never read [clang-analyzer-deadcode.DeadStores] keep_running_thread = false; ^ ~ main.cpp:22:3: note: Value stored to 'keep_running_thread' is never read keep_running_thread = false; ^ ~

I would expect no warning here since a reference to keep_running_thread is passed to the thread.

Note: You can work around this by using a capturing lambda that captures keep_running_thread by reference; but I would expect this to work using the method shown as well.

[Quuxplusone]

The warning message is actually not wrong: the code contains a data race, the compiler is allowed to replace the read operation of keep_running with a constant and remove the read of keep_running_thread.

[Quuxplusone]

Oh, interesting! I'm happy to have this bug closed then. If you're able to point me to the documentation or portion of the standard that explains this, that would be very helpful for me, just for my own learning experience. Thanks!

[Quuxplusone]

(In reply to Sam Kearney from comment #2)
> Oh, interesting! I'm happy to have this bug closed then.

Leave it open. The warning message could be improved.

(In reply to Sam Kearney from comment #2)
> If you're able to
> point me to the documentation or portion of the standard that explains this,
> that would be very helpful for me, just for my own learning experience.

[intro.races] p21 (https://eel.is/c++draft/intro.races#21)