CP.25: gsl::joining_thread missing implementation in Guidelines support library

[honey-speaks-tech]

Problem

CP.25 says "Prefer gsl::joining_thread over std::thread". However there is no implementation of gsl::joining_thread in Microsoft GSL. Hence the rule CP.25 is incorrect.

Workaround

For projects using C++20, CP.25 could be reworded to "Prefer std::jthread to std::thread". For projects using C++11 & higher, there should be a workaround to handle this broken rule.

The textbook "C++ Concurrency in action - Practical Multithreading" by Anthony Williams, Chapter 2: "Managing threads", Section 2.3: "Transferring ownership of a thread" under Listing 2.6: "scoped_thread and example usage" mentions an example implementation of scoped_thread class which could be a useful workaround to handle this scenario for projects lesser than C++20 version.

Rainer Grimm's blog also mentions about this workaround.

Action item

It would be nice to incorporate the above workarounds or a better solution to address the incorrect CP.25 rule. If this workarounds are okay for reviewers, I can support in creating a pull request to edit the rule. Let me know your valuable thoughts on this.

Thanks, Honey

Honey Sukesan

[cubbimew]

Guidelines describe what GSL should provide, implementations do their best to make it happen.

If there's interest in finalizing Microsoft's implementation of gsl::joining_thread, perhaps restarting https://github.com/microsoft/GSL/pull/582 would help, but it's unlikely, given C++20.

See also https://github.com/isocpp/CppCoreGuidelines/issues/1503 - if anything I imagine guidelines will just switch to std::jthread as part of wider C++20 rebase.

[honey-speaks-tech]

Hi,

Thank you for sharing your thoughts.

However I think a vast majority of C++ community are still not using C++20 version in their projects.

For a sample survey results, please have a look at https://www.cppstories.com/2022/cpp-status-2022/#your-input--survey

I believe C++ core guidelines should be applicable for anyone using C++14 & higher. In that case, I would be more inclined to rewording the guideline considering two specific scenarios - C++ 20 users & for others. Maintaining gsl:: joining_thread in the rule is incorrect in my opinion as it's just a wishlist & nobody is going to benefit from it.

What do you think?

[neithernut]

Well, the README says:

The aim of the guidelines is to help people to use modern C++ effectively. By "modern C++" we mean C++11 and newer.

That sentence was introduced in 2015 (back then it listed versions instead of just saying "and newer"), four years after C++11 was standardized. So I guess rewriting the guidelines for "C++20 and newer" will not happen in the near future anyway. But I'm not an editor.