boost::named_mutex remains acquired after crash

[skryv-softserveinc]

tried on macOS

[363568233]

It exists on any OS. Because the default named_mutex is spin_wait, you can view the source code, there is macro to change internal implementation. You can change it to posix pthread_mutex on macOS, now it update to robust mutex, maybe you need to update the source code.

[363568233]

Changed the posix mutex implementation to be robust #66

[363568233]

#if defined(BOOST_INTERPROCESS_USE_POSIX_SEMAPHORES)
typedef ipcdetail::posix_named_mutex internal_mutex_type;
#undef BOOST_INTERPROCESS_USE_POSIX_SEMAPHORES
#elif defined(BOOST_INTERPROCESS_USE_WINDOWS)
typedef ipcdetail::windows_named_mutex internal_mutex_type;
#undef BOOST_INTERPROCESS_USE_WINDOWS
#else
typedef ipcdetail::shm_named_mutex internal_mutex_type;
#endif
But define BOOST_INTERPROCESS_USE_POSIX_SEMAPHORES, boost using the posix semaphore, deadlock are also possible. You should use posix_mutex shared in process, but named_mutex dont't use posix_mutex. I did it myself...

[seanlis]

Would this issue ever get fixed? It has been there for almost a decade.
I spent one whole day to add named_mutex and named_condition to my windows apps, and then found out this issue forcing me to look for other solutions.

[Dalzhim]

I'm also wondering why the PR #66 was closed without ever being merged. Running into this problem is quite annoying when a known solution has existed for so long!