How does std :: chrono :: stable_clock :: now report errors?

Question

This is related to my previous question , where I asked if std::chrono::steady_clock::now should be noexcept . Now that I know I'm curious how this function reports errors? For example, a general Linux implementation of this function uses clock_gettime , which may return an error.

Answer 1

All errors that can be reported by Linux clock_gettime are not possible in the debugged implementation of std::chrono::steady_clock::now() .

Errors:

 int clock_gettime(clockid_t clk_id, struct timespec *tp); EFAULT 

tp points out of the available address space.

In std::chrono::steady_clock::now() no user tp that is not provided to the user. If std::chrono::steady_clock::now() passes the wrong tp to clock_gettime , this is just a steady_clock error that is easily fixed.

 EINVAL 

The specified clk_id is not supported on this system.

In std::chrono::steady_clock::now() no user clk_id . If std::chrono::steady_clock::now() passes the wrong clk_id to clock_gettime , this is just a steady_clock error that needs to be fixed.

Please note that the steady_clock implementation steady_clock not guaranteed to be portable and can be ported to any given platform and checked for correctness. It may turn out that steady_clock cannot be implemented in terms of clock_gettime on some platforms.

 EPERM 

clock_settime () is not allowed to set the specified hours.

Not applicable to steady_clock since there is no API to set the time.

Answer 2

This is not true. There are no error messages in the API. Thus, the Linux implementation must either handle errors internally or not return (for example, through exit() or std::terminate() ).